diff --git a/README.md b/README.md
index c1830fa177c588ef3df75240d4c5e9b4194df560..441cadfc8f66a8f5624897171ebfdccf35a3ed7e 100644
--- a/README.md
+++ b/README.md
@@ -9,11 +9,11 @@ tags:
 ---
 # Inference Kernels
 
-13 [FlashInfer Trace](https://github.com/flashinfer-ai/flashinfer-bench/tree/main/docs/flashinfer-trace) operator definitions, one per operator except for the empty-prefix case of block-causal attention, with self-contained PyTorch references, open-source baseline solutions, and workloads captured from actual inference: 11 through SGLang, and the 2 gated delta definitions from the [MLSys 2026 FlashInfer contest](https://huggingface.co/datasets/flashinfer-ai/mlsys26-contest/tree/5e832ce88dd1013032b22a0e942979d7d2769cc3). Model dimensions are constants, except where one operator serves several model shapes (weight shapes, head dimensions, windows, slot counts); those are runtime axes, like batch sizes, sequence lengths, page sizes, and image grids.
+21 [FlashInfer Trace](https://github.com/flashinfer-ai/flashinfer-bench/tree/main/docs/flashinfer-trace) operator definitions, with self-contained PyTorch references, open-source baseline solutions, and workloads captured from actual inference: 19 through SGLang, and the 2 gated delta definitions from the [MLSys 2026 FlashInfer contest](https://huggingface.co/datasets/flashinfer-ai/mlsys26-contest/tree/5e832ce88dd1013032b22a0e942979d7d2769cc3). Model dimensions are constants. Runtime batch sizes, sequence lengths, page sizes, and image grids remain variable.
 
 Every published tensor input references a complete `safetensors` snapshot. Python scalar arguments use scalar descriptors. There are no random-input descriptors or synthetic activation generators. `captures.jsonl` joins each workload UUID to its model revision, request, engine operation, input adaptations, actual outputs, and validation results.
 
-**1,104 real workloads cover all 13 definitions:** 133 Qwen Image, 384 Qwen Flash-Next, 226 DeepSeek, 207 LTX, and 154 Qwen3-Next from the MLSys 2026 contest. The 950 SGLang captures pass complete-output replay; the contest publishes no engine outputs, so its workloads are checked by running the reference. The complete dataset is published at [humanfia-lab/inference-kernels](https://huggingface.co/datasets/humanfia-lab/inference-kernels), including definitions, workloads, capture records, baseline solutions, and tensors. The 5,455 tensor paths contain 5,124 unique snapshots of inputs, complete engine outputs, and auxiliary capture tensors. Fetch them with Git LFS before running workloads.
+**1,104 real workloads cover all 21 definitions:** 133 Qwen Image, 384 Qwen Flash-Next, 226 DeepSeek, 207 LTX, and 154 Qwen3-Next from the MLSys 2026 contest. The 950 SGLang captures pass complete-output replay; the contest publishes no engine outputs, so its workloads are checked by running the reference. The complete dataset is published at [humanfia-lab/inference-kernels](https://huggingface.co/datasets/humanfia-lab/inference-kernels), including definitions, workloads, capture records, baseline solutions, and tensors. The 5,506 tensor paths contain 5,124 unique snapshots of inputs, complete engine outputs, and auxiliary capture tensors. Fetch them with Git LFS before running workloads.
 
 The dataset was regenerated from full pinned checkpoints through SGLang on NVIDIA B300 GPUs. Definitions, reference code, tolerances, and baseline solutions are unchanged, apart from the [curation](#curation) for kernels that real inference uses.
 
@@ -23,23 +23,23 @@ BF16, FP16, FP32, and FP64 mean 16-bit Brain Floating Point, IEEE half precision
 
 | Family | Model | Fixed dimensions | Definitions | Workloads |
 |---|---|---|---:|---:|
-| Dense matrix multiplication | Qwen Image 2.1 Turbo | `(N,K)`: `(4096,4096)`, `(12288,4096)`, `(4096,12288)`; `M >= 4096` | 1 | 93 |
-| Grouped expert matrix multiplication | DeepSeek V4.1 Flash | 384 experts; `(N,K)`: `(4608,5120)`, `(5120,2304)`; FP8 × FP4 | 1 | 104 |
+| Dense matrix multiplication | Qwen Image 2.1 Turbo | `(N,K)`: `(4096,4096)`, `(12288,4096)`, `(4096,12288)`; `M >= 4096` | 3 | 93 |
+| Grouped expert matrix multiplication | DeepSeek V4.1 Flash | 384 experts; `(N,K)`: `(4608,5120)`, `(5120,2304)`; FP8 × FP4 | 2 | 104 |
 | Mixed-format latent attention | DeepSeek V4.1 Flash | 64 heads × 512 dimensions; 128 FP8 window slots and 512 FP4 compressed slots | 1 | 42 |
-| Sparse multi-head latent attention (MLA), prefill | DeepSeek V4.1 Flash | 64 query heads × 512 dimensions; 1 shared key/value head; BF16; 128 or 640 slots | 1 | 30 |
+| Sparse multi-head latent attention (MLA), prefill | DeepSeek V4.1 Flash | 64 query heads × 512 dimensions; 1 shared key/value head; BF16; separate 128-slot and 640-slot definitions | 2 | 30 |
 | Gated delta recurrence | Qwen3-Next (MLSys 2026 contest) | 4 query/key heads, 8 value heads (TP=4 shard), dimension 128; k-last state; prefill and decode | 2 | 154 |
-| Noncausal dense attention | LTX-2.5 | 32 heads; dimension 128 (video) or 64 (audio); no mask | 1 | 162 |
+| Noncausal dense attention | LTX-2.5 | Video: 32 heads × 128; audio: 32 heads × 64; no mask | 2 | 162 |
 | Qwen sparse attention | Qwen 3.8 Flash-Next | 24 query heads, 2 key/value heads, dimension 256; 512 four-token blocks | 1 | 192 |
 | Prefix-cached block-causal attention | Qwen Image 2.1 Turbo | 32 heads × 128; empty-prefix and cached-prefix cases | 2 | 40 |
 | Sparse index scoring and selection | Qwen 3.8 Flash-Next | 4 query heads × 128; Top-512 complete four-token blocks | 1 | 192 |
 | Expert router | DeepSeek V4.1 Flash | Hidden dimension 5120, 384 experts, Top-6 selection | 1 | 50 |
-| Three-dimensional neighborhood attention | LTX-2.5 diffusion decoder | Dimension 64; `(heads, window)`: `(32,3×7×7)`, `(16,3×7×7)`, `(8,3×5×5)`, `(4,11×11×11)`; the window is an input | 1 | 45 |
+| Three-dimensional neighborhood attention | LTX-2.5 diffusion decoder | Dimension 64; `(heads, window)`: `(32,3×7×7)`, `(16,3×7×7)`, `(8,3×5×5)`, `(4,11×11×11)` | 4 | 45 |
 
-**[Shape distribution histograms for all 13 kernels](docs/shape-distributions/README.md)** show how the 1,104 workloads are distributed across exact runtime shapes. Each bar counts captured workloads with the same complete combination of variable dimensions; fixed model dimensions are listed above the plot. These counts describe this dataset, not AgentX traffic frequencies. Regenerate the figures with `python scripts/plot_shape_distributions.py` (requires Matplotlib; no GPU or tensor files needed).
+**[Shape distribution histograms for all 21 kernels](docs/shape-distributions/README.md)** show how the 1,104 workloads are distributed across exact runtime shapes. Each bar counts captured workloads with the same complete combination of variable dimensions; fixed model dimensions are listed above the plot. These counts describe this dataset, not AgentX traffic frequencies. Regenerate the figures with `python scripts/plot_shape_distributions.py` (requires Matplotlib; no GPU or tensor files needed).
 
 Definitions describe complete logical operators. Captures preserve all model heads and experts; distributed shards must be gathered before publication. Backend padding, layout changes, and page remapping are recorded explicitly. Cache snapshots retain every referenced page, including its unused tail; unreachable allocator pages are omitted. Tensor strides are normalized to contiguous storage without changing values.
 
-The gated delta definitions are the MLSys 2026 contest's: raw Q/K/V with the gate inputs `A_log`, `a`, `dt_bias` and `b`, and the k-last state layout `[sequence, value head, value dimension, key dimension]` that serving engines keep. Prefill takes variable-length sequences through `cu_seqlens`.
+The gated delta definitions are the MLSys 2026 contest's: Q/K/V with the gate inputs `A_log`, `a`, `dt_bias` and `b`, and the k-last state layout `[sequence, value head, value dimension, key dimension]`. Prefill takes variable-length sequences through `cu_seqlens`. These are unchanged Qwen3-Next TP=4 contest contracts, not a layout conversion of the removed Qwen Flash-Next captures. Their references do not perform Q/K L2 normalization, and they accept per-batch states rather than the original SGLang state pool and slot indices.
 
 Quantized byte carriers use `int8` because the trace schema lacks unsigned-byte and exponent-only scale types. The byte patterns remain unchanged. Grouped matrix multiplication uses logical scale layouts with one scale per 32 K elements. Mixed attention preserves the engine's physical pages: each page stores all payload rows followed by all scale rows. Both cache formats share the softmax denominator, including the zero-value attention sink.
 
@@ -104,7 +104,7 @@ The existing baseline implementations and B200 measurements come from the origin
 
 | Kernels | Open-source implementation |
 |---|---|
-| Dense and grouped matrix multiplication | [DeepGEMM](https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d) |
+| Dense and grouped matrix multiplication | [DeepGEMM](https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d); selected small dense shapes use [FlashInfer](https://github.com/flashinfer-ai/flashinfer) 0.7.2 CuTe DSL |
 | Sparse MLA prefill and mixed-format MLA decode | [FlashMLA](https://github.com/deepseek-ai/FlashMLA/tree/0f11e178da1f67becb087a324b400854931ec3e3) |
 | Dense, block-causal, and sparse grouped-query attention | [FlashAttention-4](https://github.com/Dao-AILab/flash-attention) 4.0.0b35, with runtime masks or sparse-input adapters |
 | Gated delta recurrence | The MLSys 2026 contest's FlashInfer prefill and decode wrappers, unchanged |
@@ -130,6 +130,8 @@ The runner uses the native Trace builder, validates every output element against
 
 Each SGLang-captured definition also has `solutions/<family>/sglang_<definition>.json`: SGLang's own implementation at the capture commit `684fea8a`, called the way the engine calls it, so that a kernel can be timed against what serving actually runs. The module docstring cites the engine call path. Run these solutions inside the capture image `lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8`. Each module also exposes `prepare(*inputs)`, which does the host-side setup the engine does once (layouts, block masks, workspaces) and returns a zero-argument `launch`; `run(*inputs)` does both.
 
+The SGLang solutions adapt the dataset's independent operator interfaces to the pinned engine implementations. Those adapters include padding, layouts, per-rank calls, metadata and output ordering where needed; their module docstrings describe the boundary. A passing standalone workload does not establish that a candidate can replace an engine call without integration work.
+
 ### KDA acceptance configs
 
 `kda/<definition>.json` is how KDA (Kernel Design Agents) judges a candidate kernel on that definition. The definition's reference is the truth, and the SGLang baseline above is timed live on every workload, with only `launch` timed. The fields are:
@@ -139,16 +141,12 @@ Each SGLang-captured definition also has `solutions/<family>/sglang_<definition>
 
 ## Curation
 
-This revision keeps only kernels that real inference uses, as KDA (Kernel Design Agents) tasks require; a kernel written against a definition should drop into an engine.
+This revision curates inference operator tasks for KDA (Kernel Design Agents). Per-shape definitions, workload files and solution names stay separate; model dimensions remain constant axes. Captured interface adaptations are retained explicitly.
 - **Removed `dense_gemm_bf16_nt_n64_k4096` and `dense_gemm_bf16_nt_n4096_k64`.** These are the image output and input projections: memory-bound 8–45 µs GEMMs that run once per forward.
 - **Removed `grouped_gemm_fp8_fp4_e384_n2304_k5120_g32`.** It was cut from the columns of the concatenated gate/up GEMM; no engine runs a separate gate or up GEMM.
-- **Replaced the gated delta definitions.** `gated_delta_rule_{decode,prefill}_h48_dk128_dv128` gave Q/K pre-repeated to 48 heads and a `[B, H, K, V]` state, which is not an engine's interface. They are replaced by the [MLSys 2026 FlashInfer contest](https://huggingface.co/datasets/flashinfer-ai/mlsys26-contest/tree/5e832ce88dd1013032b22a0e942979d7d2769cc3) definitions `gdn_{decode,prefill}_qk4_v8_d128_k_last`, with their workloads and FlashInfer baselines unchanged. Each input tensor is stored in its own content-addressed file. Each receipt records the source workload, plus a check that the reference returns finite outputs of the declared shapes.
+- **Replaced the gated delta definitions.** `gated_delta_rule_prefill_h48_dk128_dv128` pre-repeated 16 Q/K heads to 48; the old decode kept 16 raw Q/K heads and included normalization and gating. Both captures extracted states from the engine pool and transposed them to `[B, H, K, V]`. They are replaced by the [MLSys 2026 FlashInfer contest](https://huggingface.co/datasets/flashinfer-ai/mlsys26-contest/tree/5e832ce88dd1013032b22a0e942979d7d2769cc3) definitions `gdn_{decode,prefill}_qk4_v8_d128_k_last`, with their workloads and FlashInfer baselines unchanged. Each input tensor is stored in its own content-addressed file. Each receipt records the source workload, plus a check that the reference returns finite outputs of the declared shapes.
 - **The three remaining dense GEMMs keep only their `M >= 4096` rows.** The `M <= 32` rows were 11–27 µs launches.
 - **The LTX dense attention definitions drop the dense `[B, H, Nq, Nk]` visibility mask.** It was all True on every captured call (up to 839 MB). On one GPU the engine passes no mask, or an all-visible prompt mask. Replay against the captured engine outputs still passes.
-- **One definition per operator.** The per-shape definitions of an operator are merged into one, in which the shape constants are runtime axes: dense GEMM `N`/`K`, grouped GEMM `N`/`K` (with `K_packed`, `K_groups`), dense attention `D`, neighborhood `H`, and sparse MLA `TopK`. Block-causal attention keeps its empty-prefix definition (`Np = 0`), because FlashInfer Trace requires positive runtime axes; KDA judges the two as one task. The references were identical across shapes. The neighborhood window becomes three scalar inputs (`Kt`, `Ky`, `Kx`), because no tensor shape carries it.
-  - Tensors move to the merged definition's directory; identical snapshots collapse into one path each.
-  - Receipts take the merged definition and axes, and keep the definition they were captured under as `captured_definition`.
-  - The per-shape open-source solutions merge too: one DeepGEMM BF16 solution for every `(N, K)` (the CuTe DSL branch served only the removed `M <= 32` rows); one DeepGEMM grouped solution; FlashAttention 4 without a mask (FP16 for `D = 64`, as before); NATTEN with the window from the inputs; and one FlashMLA solution.
 - **Receipts and tensors.** The receipts of removed workloads are dropped, along with every tensor that nothing references.
 - **Scripts.**
   - `check_workload_inputs.py` evaluates constraints that read inputs, such as `cu_seqlens[-1]`.
@@ -231,7 +229,7 @@ BF16 comparisons use `rtol=0.016, atol=0.005`; FP32 uses `rtol=1e-4, atol=2e-5`;
 
 One complete router workload is excluded from the 279 DeepSeek raw captures: at two rows of a 4,096-row call, FP32 ranking ties produce different expert selections. The original FP32 reference and exact selected-set requirement remain unchanged. This call was also absent from the original dataset. No rows are removed from published tensors. All nine LTX requests pass and are included.
 
-Tensor files are published directly as `blob/workloads/<definition>/<hash>.safetensors` through Git LFS. Every workload and capture record references tensors under its definition directory. Shared snapshots have identical LFS objects across directories: 5,455 paths contain 5,124 unique snapshots totaling 146.34 GB (146.41 GB when materialized at every path).
+Tensor files are published directly as `blob/workloads/<definition>/<hash>.safetensors` through Git LFS. Every workload and capture record references tensors under its definition directory. Shared snapshots have identical LFS objects across directories: 5,506 paths contain 5,124 unique snapshots totaling 146.34 GB (146.44 GB when materialized at every path).
 
 ```bash
 git clone https://huggingface.co/datasets/humanfia-lab/inference-kernels
diff --git a/blob/workloads/dense_gemm_bf16_nt/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors b/blob/workloads/dense_gemm_bf16_nt_n12288_k4096/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n12288_k4096/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k12288/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k12288/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c463d3327cbf1807b2c0c239f01e95d5e3889155
--- /dev/null
+++ b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc80be89d826bc381aa7857a75cf96f56eb74ba618140579ced9040862e219b6
+size 33554520
diff --git a/blob/workloads/dense_gemm_bf16_nt/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors
diff --git a/blob/workloads/dense_gemm_bf16_nt/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors b/blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors
similarity index 100%
rename from blob/workloads/dense_gemm_bf16_nt/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors
rename to blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..087b47549cb274355b8a150a9d186ede04872be1
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c010c9cf19b75e96beeec58d1dd871f737e58b6a4a469a10f986cd2a652aa354
+size 47696
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..a81a9bb65a7e6eaf5b9be3ce98890d84670fb9d5
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cec503160c09bfde2afa8d4fa55701fbbc3bfb22b2c8916bd83c72be27259ab
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..f1fda3d3d698eabc7b09aa1ae39f458f7ea61dd2
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25b51d2f6d79ed6a78746073713ba27f46905a3af26e3684d12e19ee4bc12aac
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..bd50736d0898c30e586ba571e0d83be624bf92e3
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec4b5545494dd3f963226444f44981e76d4449b31aa040355ab66b7bdeaa1945
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..1c04a94291b918c036999310c39eaf10261c7772
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a0330707305c42493a3f6826265442e28ef98c50ccd4c24523043df64456f16
+size 83024
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c51dc76c4cca3a547550065f9afd7fd3f7d531e6
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f45a2af45a7772bee6fdad9eb2dd71cc7c9cc1c5b4f4c7ea6f97b8b7e64433e4
+size 120
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..03e19da9cd58226729b6af3e7d9069a4112b25d0
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:513a6cdde4e6fede8e8d4220673dcaa872938c05b0692e3f1e594068dac77bd8
+size 98384
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..a16cd7015d9792cb0ab9e0a81a99da6bcfc0663c
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c929361b039d901f65a20d7d15cd8a1865bb8f05b35994e454d683af5504eef7
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..903ec6a8f9bfbfdb038c5858f50bf4869b735ff1
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d190efeb7efbab0c48735c17a24b616f987de6b3193c6a424b31356f67da091
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..4900be94d9cb80b99c7342f2d18da4ec00651b89
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81fc011172fa2e0331809ba464aea06c25df0578f5dd9ab67784c00bf7dfa916
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c2687a2498bd61b9a5a47b13340436765a875881
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f6a8b0038db7729aa1a93a9345a2d35b9928d21a7e7426730efe81bf9608b35
+size 3144
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..2cf721463adc7552ff5275c2bf167e50dc5b75c2
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab4ab9d85b7902c4e4fbcf5ddcd046c2934602144115b30e19f096737c8efc46
+size 98384
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ce0536c2723d145cba71da6486c80f96abc63116
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1df43feffbde19463477d74b951f23591a2e0c796a7a9575789df19201478bb3
+size 120
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..547def4caa5b455f11bb2c00ff5774252e0aaaf5
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4819319a740b82945a4a8233dd8cd94447e9b1aaa84226d6055b546bf7b0ced
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..93e0d912508f69b747776608d7bb75aedaf06a87
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:970bde7a522af3cffc676aafb0f765b65b3f35fab2e5172131087800ffc1142f
+size 144
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..6cba0252971921a317be269b9271f306877e50e2
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90a56a74cfc3dfb08042bc3000fe4996320eb3fac5cca3bb5b49e9d412c1d239
+size 120
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..634014ea12cb17bb134fe2b891bfcc5bbc674ba8
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc802716d5e1c7a424a1a363dda944833962a07e60966372f12ff0ec8e113856
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..49678a3e51554cff1b17b7c9d82b7bb4c17d242a
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5adf574aa4f581db584e0d38448c7a3b331574e4262c5f12293055320af1696d
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..42109f8e3257726cdc9209207595e135a6f23022
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf4540bdf4ea2941723541709afcdfce88b73fc696b6436f9ba540beccc5aa3d
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ace75679a935f3f2d8e66134d76501b75e3721d8
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb9c4ff6556c90c235b82da37e58ff4b6477e06d84b0f5b8da6d03c1ae8ad2aa
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9876b5b1a1757b4baec21f692d30644d2d6ace50
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c943f36ccf9ff3d47dafa92cfcba0309ffa7baf10d411f227f1943076da43ff
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..ef5de55cb9e5fb77431c0b8f65a28351c7d02be9
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fc6b7c380baf2c6603dbeaeac622f45effd5a2832ba798110d24f67c02cef2e
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..713c15eb5c6dec84a1c75864bd3139c1634c5bbb
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:170de549c10205507d2c7dbe2e86af3c949fbf949b5449f65fba337842ce19f1
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..4c83176aebf4d31c4f0c146f03d1e28d5052e1be
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bcf835db256840cefc39436b6ee4dfd6a01d74ea289efc122a4586dcf98d890
+size 144
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e8b3b22bddff7048abf773d266850a7ed7e624b6
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8411aeddfb3a00aa7225d57d03d50d1fd6c4ea24f11fc0caf1a4a588e5ca751
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..60c658212df37b5eb62d7ea8b836b91078d496e2
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f32ddc95f5da2c8766ea4125258f94111100314c25b460f8ef9ae0fa412c5f6
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..307136cf068e07894fa409843d6d07d9c0003ac9
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4be3dbcd4bf648db8c5cec10ff1b522ba7ca4d606429d5e9e54f0d41da654bb0
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..74dbaae72b7e184432e2999b77a0177b1643c46e
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eb96b1fa5c363fbc2b6442a8993669d79f502cdd3af25c2b3d4bb2aff753e08
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..423f4e4eaf2b25c5024da18f7d904fec57cfd118
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bf8e31f20ec67bcbd6981bbf56b25341624d535a0eb854799ea53e99496df2a
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..5963aa8100a01989b87bc57b0e025bfe5f02382d
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:845063acdd9a2cd0714cb97b3310f822d12928850ab5cbda2dc79a3c70fd9e0b
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..7c859c3fdfd44d431ed7a9c9733131d7210880ec
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f1d76f49ab4e57bd17eabf91469fb008595fe2e671d8d747a7a71ef2520128a
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..088567e37562a0760779662d79ddcb9fbbfb05e3
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5544acba28639282652f478d549cea4b1153a0ba55e4d6c396563b3a2396042c
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..1914142b8997ebe5fec66c4624d61f03a2ffa64e
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46b9634b7abce01f4ccb8891e5d154ea608d3381c4fe8edaadc91e83955c4970
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e4c4e39cdc6f544234532a164aa9223c6165afd4
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18938eb3eba7deccd7267fd27288a80cf584050c9e0a3dcdaa9874ab240d0a87
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..9c021aef37cf8c42beeba4b8384448adfb2e799b
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6644c0517974c921c88a0c07ad310dc944790f59fa89db3ea60744748281f06d
+size 120
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c0a48673bae612f627ab5c6421884bedceedf1bd
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4e62cfc82f0a50a4b09a740d52c17af0c9760311e1a594c204981438027440a
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..38101b04959d17c35795086949996efd29857e5d
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:499e60f024902d97faf2ddb6e0aa5431bab68a07700b2a1d8fa4d5dd0c725248
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..e9e009c7f3ea8ab4d25c1344e0093265072d0c2e
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d76df2b7ec5a96b10dc1ed1ab41cd1d92fc4b7d7994b15fbaeba861d5a635ebb
+size 70736
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..da6a3b5ad17c120c03c7aef368d704eba9e1913c
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1278148af40a6c1e2ae123acb4db47581f58210bb3420951413b8181c9edd3d2
+size 144
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..b4a1282293a32dd69346297bb67cd00f71b414d9
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3308f6740912afea7823c5262a5aa3c4f40fc6b897b131f10a494fd2639ae5cf
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..62e59816b4d1a63729f1e3065058a71b323e99d7
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f000c4662617d197aeea29836a18456216c54211df68613ba10e3ee5e7911cf
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..f4ac416f762648d5bc592e11c532379fe4a30ce5
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a2f516e432d5d45937125180e26a823c16ab0acab111c6cb8a53de03e88b5fa
+size 96848
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..488bc40c5007c796199d9db0d3503e89672bd17a
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13dea4212417e72f42588eb481797703a2de8d51e36667b1f5a4265447b1fb8a
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..1103c1a3d951b875db50b7ed74573308e63b9ddb
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:375f26d6f9225afaeccf8bc28a89e6d3a4e7c11630040b834cc6cbbb38cb950b
+size 58448
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..7cb620bf842cd6a8d57806558dd009ebb5438086
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b25fed5af2ac1da13d58d5cfc218934dd26f22fcd838765b7d53c13e98e92659
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..0c6cc934148ca6a7486371482077bd7072520577
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e938580fc591adc05d1dfd70dbecabc9402f4983d26b859447bde654ffa0b7bf
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..3b09ce3ce357cbdbcd29ad33e37de1657ef1e295
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8fe880ca0cdab1566cc63adab0da4d34927d0fc7ba4310504836b778e92825f
+size 144
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..2a977cbccf8ecbbc5198da660269935f22989e88
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15a8503524b1ca2661d68182668965e109f9681d2ff2af0bc917e9651c956dcf
+size 98384
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..98db23b7187b3c76bb205efeb6063e377b18c122
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce0ba0eeabc251c1519afd2cd08dfeb60ec13d9d266376ee340938898068e9a8
+size 10832
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..59e655b7b3a61c2eb6a471884ade482c0d9f8c81
--- /dev/null
+++ b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28ca5204d5dcd35d400a24c4970acc1a0fd7b8b5fcfbd7d07edacfc3703647f3
+size 96
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors
diff --git a/blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors b/blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors
similarity index 100%
rename from blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors
rename to blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors b/blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors
rename to blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors b/blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors
rename to blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors b/blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors
rename to blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors
diff --git a/blob/workloads/neighborhood_attention_3d_d64/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors b/blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors
similarity index 100%
rename from blob/workloads/neighborhood_attention_3d_d64/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors
rename to blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d128/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d128/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors
diff --git a/blob/workloads/noncausal_dense_attention_h32/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors b/blob/workloads/noncausal_dense_attention_h32_d64/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors
similarity index 100%
rename from blob/workloads/noncausal_dense_attention_h32/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors
rename to blob/workloads/noncausal_dense_attention_h32_d64/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k128/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k128/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors
diff --git a/blob/workloads/sparse_mla_prefill_h64_d512/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors b/blob/workloads/sparse_mla_prefill_h64_d512_k640/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors
similarity index 100%
rename from blob/workloads/sparse_mla_prefill_h64_d512/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors
rename to blob/workloads/sparse_mla_prefill_h64_d512_k640/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors
diff --git a/captures.jsonl b/captures.jsonl
index 134609755fad3c92beb4fb48e0aa1bd9e29a5fad..ef3fb0e40f3d985d56f90e93609d6cf2b999e476 100644
--- a/captures.jsonl
+++ b/captures.jsonl
@@ -1,343 +1,343 @@
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.12174165714532137, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.024486311711370945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.12174165714532137, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/572a11dae6f7793232f1ecc6d7792779dbd95e0ecbf87877fef7f37fd296deaa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.024486311711370945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/efa0d870b83867f8ea856ba39ba809b330fca342415994531f7ecb7529d799ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "81e94812-2329-5af8-ae49-8167cf27ca32", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 14.591754162218422, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "81e94812-2329-5af8-ae49-8167cf27ca32"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 25, "Nq": 25}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/ec61a985467f7c12f4496197b00f64475ab6ac5f5b4e142527235e7ae823a9d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 25, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "361da807-31fa-5a25-8b62-37752b8f3529", "validation": {"axes": {"B": 1, "Nall": 25, "Nq": 25}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.008140245918184519, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "361da807-31fa-5a25-8b62-37752b8f3529"}}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05146070662885904, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.04754280112683773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2c87c6952f0b539065032448466ba4e943b4e5892d90c8b3a67b5f379a140233.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 0, "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05146070662885904, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c73097f491d3f526d743389cd71d1c69f174db8d2d1030ab6d6fa3fd8a0c606e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.04754280112683773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/398b0e6f0079beab22d824bd2d7732d057a63eab0bf56882656c29eb5bf6dbcd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "6dbb2b24-0ece-59ae-b07f-05a11948c833", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04182086419314146, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6dbb2b24-0ece-59ae-b07f-05a11948c833"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02600828092545271, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.04886566987261176, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.047110469080507755, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e432defa3011150db18920d77084699d442a88e2349d4234518b7993a045960e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02600828092545271, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5c56ec3d96e05c1d64042ed96d3262d2660d28b20330141694069ac6c8af13d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 4, "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.04886566987261176, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3c7c50a338fb849474ab1ac4d6e12b6892ae5c16050ad027fa3bb2dd41ea297c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.047110469080507755, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/29fa773ea9d4948cbb1d8d03edff83594e9f28d4fdc52ee3186bb05e9b3645b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "aa1790c4-e4c4-53e6-9e27-0dd837883353", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03917623497545719, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa1790c4-e4c4-53e6-9e27-0dd837883353"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02400219812989235, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "1ef85b69-b938-5554-b019-1cd333110897", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.0463019828312099, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1ef85b69-b938-5554-b019-1cd333110897"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 6144, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc", "validation": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.18826853577047586, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c10722b8773a8ec6586cdf8a3722cb5480a25a24a44fcad1784d1797dfa473f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02400219812989235, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/31f9f63df8a657253651b29f9dad25de8a5b00883933c1e9d1006f238cd217c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239230.3971937, "step": 7, "uuid": "1ef85b69-b938-5554-b019-1cd333110897", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.0463019828312099, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1ef85b69-b938-5554-b019-1cd333110897"}, "variant": "rows_4096"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0e8a0eb1b0bbd588a87c0110b5b93be119538d4a458790c6de2449c22c941159.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.18826853577047586, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc"}, "variant": "rows_6144"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/164724b8f736981258f69907c85c883cf083dc5f393900970197af72cc45da0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "dc3b6e8c-6201-52c8-9d2e-a190d1562087", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.10301753412932158, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "dc3b6e8c-6201-52c8-9d2e-a190d1562087"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 30, "Nq": 30}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/5182ae66ea455594116b0c41913cb8f9d94a2fedcad7e5a5b4d58b09ca2a88f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 30, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "3a45e86c-e517-57d7-92c0-85586056d271", "validation": {"axes": {"B": 1, "Nall": 30, "Nq": 30}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.004042515996843576, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3a45e86c-e517-57d7-92c0-85586056d271"}}
-{"axes": {"K": 4096, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "bc334c45-f866-5289-891d-4e12672c7e97", "validation": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.03487044433131814, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc334c45-f866-5289-891d-4e12672c7e97"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60", "validation": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07722086180001497, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 6144, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505", "validation": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.06412861868739128, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f5441516577f6bd6bdd40b2d516564506a857374b1be4dd6dde0bf475febf2cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "bc334c45-f866-5289-891d-4e12672c7e97", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.03487044433131814, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc334c45-f866-5289-891d-4e12672c7e97"}, "variant": "rows_6144"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d01f888681bf35341b734854873d23d2baee4810670db420f469f27f1f244aff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 0, "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07722086180001497, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60"}, "variant": "rows_6144"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/862b917b34836b4cb5c914ebd438468ab52cde111f8fd09bc92099d58c6d67be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.06412861868739128, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505"}, "variant": "rows_6144"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/c59769297a99e59f2b7c707af5d819e62dc0c4efe7320925aead09f72af4d6e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "2b9361ff-dbd3-598b-95fc-eeec388cc422", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06502548418939114, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2b9361ff-dbd3-598b-95fc-eeec388cc422"}}
-{"axes": {"K": 4096, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886", "validation": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.031620386987924576, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1", "validation": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.0747010437771678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 6144, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f", "validation": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.0760988718830049, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/2b48954b969eecec2384645c15d37f3a6bb884cd33ecc45d78fc12488e89dcf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.031620386987924576, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886"}, "variant": "rows_6144"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/809d2b82c51d3f7e33329ebf2c4e60e496995095d76f613140e91a9887feae39.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 4, "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.0747010437771678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1"}, "variant": "rows_6144"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f198676a629599e54864f76f9630326c23ab09bcdd94e89e26445da6898c418.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.0760988718830049, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f"}, "variant": "rows_6144"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/a5bfe0abdc234ffa77ca9d6fd583bfa6de4e7f6f921a4aa6aef800ac3e17c910.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "0d5a8062-d653-5f8d-89e7-c92004279f8f", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06466197408735752, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0d5a8062-d653-5f8d-89e7-c92004279f8f"}}
-{"axes": {"K": 4096, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5", "validation": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030148770194500685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 6144, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57", "validation": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07489119237288833, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57"}, "variant": "rows_6144", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.042013777419924736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 4096, "M": 4125, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260", "validation": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02451479807496071, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ee2964bf2018a2348bd2142f30f108e23d5a838b247cb26e7f86799ba00910ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030148770194500685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5"}, "variant": "rows_6144"}
+{"axes": {"M": 6144}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b43cd19bb9d3d07d9314264664cce99e0c13c5a7e5a892ca9531c5a4fa5e5cfc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239239.1825383, "step": 7, "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57", "validation": {"axes": {"M": 6144}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07489119237288833, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57"}, "variant": "rows_6144"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.042013777419924736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48"}, "variant": "rows_4096"}
+{"axes": {"M": 4125}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.02451479807496071, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260"}, "variant": "rows_4125"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/133f657a23877ad38fb6e1a13cfe79fca69d3c0b0a42041db5e61af4de8100c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "322fac48-fa66-5a79-aa96-a99c862d24e2", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06613255385309458, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "322fac48-fa66-5a79-aa96-a99c862d24e2"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/88d8d74e04b85e837291f3803e67f4fcc0c0f34e2c657e377188c36be21a3217.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "f2a1a8e2-460a-549a-a3d1-bdf74bf42f26", "validation": {"axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.03569585597142577, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f2a1a8e2-460a-549a-a3d1-bdf74bf42f26"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.023155256174504757, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.04904062207788229, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4125, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "717bab44-772a-580e-bd31-8f449703fca5", "validation": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05703906808048487, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "717bab44-772a-580e-bd31-8f449703fca5"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 4125, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1", "validation": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.05304816411808133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.04559852974489331, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/00b199614f81d11ba2763b59714d61372a4c2add5d193d6e896b191f280974e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.023155256174504757, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/6456dd42a050590735972eea6e10edba0fc9b03e040bfda0724a0402340eecf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.04904062207788229, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a"}, "variant": "rows_4096"}
+{"axes": {"M": 4125}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "717bab44-772a-580e-bd31-8f449703fca5", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05703906808048487, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "717bab44-772a-580e-bd31-8f449703fca5"}, "variant": "rows_4125"}
+{"axes": {"M": 4125}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 0, "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.05304816411808133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1"}, "variant": "rows_4125"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/6e4dd7f63ee604011aba960400d128e5ed7b7bb21e8e6869be5ee2c10cf44be2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.04559852974489331, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/76201a82d4c5fc6b8643ae7f6a38c9007047cab851982ac1f6239fb9d4b4f354.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "76691d76-05ea-597b-a3d5-147ef371c820", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06718758028000593, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "76691d76-05ea-597b-a3d5-147ef371c820"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026999427005648613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.045856614131480455, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.05973345506936312, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/51bd9c5c57d67f6e9415a85862a0eff814b53655c3f6ed655113b998da86427f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026999427005648613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/526c870125480519a48c9a638517ada4566b78891bcc40259acd566429c1ee30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 4, "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.045856614131480455, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/15b7d5bdf696f9737135f2b0fb9c070e7b7c314a47f94d42f2b7f2220b463f08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.05973345506936312, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/7ed73de575477d876ee2981ece7acd672789ba6fb09b3b83b214de5a92cb7ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "db82b24b-937b-5711-9558-e521b593d474", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06704403925687075, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "db82b24b-937b-5711-9558-e521b593d474"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026603277772665024, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.05827482184395194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.03990104701370001, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/934b83cf86460e9b9742b93a6bb81dac6357bd596b3ccc7dc262b8b997166892.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026603277772665024, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c931b56df8f13f62bc06c41dd9e86900fd4976b4045c68fe113847a21699d288.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239246.744593, "step": 7, "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.05827482184395194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dc9b20c2bc55f03a1dd7c18d929a141cc71ef4cbca2e2013c7ffa2642e1fa333.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.03990104701370001, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/10717ae953e6012412024c6acb8b0875873bb08ac3b6ffd0754feb989dd839ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "b1e89f95-4b54-5116-9a57-3c29fb4cb095", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04190012393519282, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b1e89f95-4b54-5116-9a57-3c29fb4cb095"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 32, "Nq": 32}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/16b2415f217ef91a9a67466dc67b4f5cf097d249377e14a464187e8006435bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 32, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "d90f0f73-d020-5734-8dd7-710a91a624a7", "validation": {"axes": {"B": 1, "Nall": 32, "Nq": 32}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.003915837965905666, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d90f0f73-d020-5734-8dd7-710a91a624a7"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.024482560344040394, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.058263967745006084, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.0601740600541234, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/92fe8c873d3a111e4563dbce2879fd6db88a4b76f03ca3534d44990d8b28f2af.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.024482560344040394, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2f7dcef704d0c2fdfcca395b1fe0a176c03e42955129ebb86e6723dfe76e646b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 0, "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.058263967745006084, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5c0f576a693acfced57412c7695e5a49841a899f035f24edc61c2d0696aee7a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.0601740600541234, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/0113132572dd13083893beeb5b5b233cd47fe3d90a64624d717e848ca60f10da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "9bdf7136-45be-561a-9076-36477832e84d", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04426139825955033, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9bdf7136-45be-561a-9076-36477832e84d"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026707913260906935, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.061376822646707296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4096, "N": 12288}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c", "validation": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.05842510890215635, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/de7120ff63282d8cd61da8fa938b9b21402af820bd1b03277d666220a548cf35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026707913260906935, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f238a5250b37befcbc6e190fdeccbc6628166ecccf2c5bba5914dba4758f168e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 4, "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.061376822646707296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/10421991d48239526ec8f05a46c718c16d90cacd498a0ffb4f4c60c1c3d61f3f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "torch.mm", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.05842510890215635, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c"}, "variant": "rows_4096"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 4, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/e1530ed27fbd9f4871ae26741dcb8a00be93a96ab263ece6a72bb327ec06e2c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "cce65778-0b96-5f2a-8a7b-16a26e8a2425", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.044299577828496695, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cce65778-0b96-5f2a-8a7b-16a26e8a2425"}}
-{"axes": {"K": 4096, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "2a724567-1566-5652-ad8c-2546de654e3c", "validation": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026511874049901962, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2a724567-1566-5652-ad8c-2546de654e3c"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 12288, "M": 4096, "N": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33", "validation": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.06076050875708461, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33"}, "variant": "rows_4096", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.030111619271337986, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/18a22a7fae925dc56b2b3c1c43f379196fbfc23e7a31766ef0c7958e4f74b322.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_out.0", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "2a724567-1566-5652-ad8c-2546de654e3c", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.026511874049901962, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2a724567-1566-5652-ad8c-2546de654e3c"}, "variant": "rows_4096"}
+{"axes": {"M": 4096}, "capture_steps": [0, 4, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/473a0e2997cbabadc85352895a553f7f9a824a67b08132b7309acdd1c14167d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 1, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 588, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 1, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha", "requested_outputs": 1, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b1-r1/results/images", "--input-save-path", "/task/raw/image-b1-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239254.438051, "step": 7, "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33", "validation": {"axes": {"M": 4096}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.06076050875708461, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33"}, "variant": "rows_4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.030111619271337986, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8"}, "variant": "rows_8192"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/efa0d870b83867f8ea856ba39ba809b330fca342415994531f7ecb7529d799ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "0fded79d-d450-5a30-a52a-054d2aba0b5e", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 20.078923034016043, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0fded79d-d450-5a30-a52a-054d2aba0b5e"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 25, "Nq": 25}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/ec61a985467f7c12f4496197b00f64475ab6ac5f5b4e142527235e7ae823a9d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 25, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "8d248f2d-1847-5526-a70d-1382a83477f8", "validation": {"axes": {"B": 1, "Nall": 25, "Nq": 25}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.006427989806979895, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8d248f2d-1847-5526-a70d-1382a83477f8"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.24693757621571422, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05861895019188523, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.027585350442677736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/224e353933f43af734ec74834219cd0d532f8a4c2a5161f1bd6cb684e9f0c653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.24693757621571422, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9fa10811c2bd2c54d53afbac8b7c887ca75deec828a11123642283d5266b98e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 0, "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05861895019188523, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e9a32d3ed18cffbada4d9c8f2ecca15565f466ed74fbacaa98f9d5b72bf7137a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.027585350442677736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979"}, "variant": "rows_8192"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/29fa773ea9d4948cbb1d8d03edff83594e9f28d4fdc52ee3186bb05e9b3645b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "7eb997b9-50e5-59e0-8a3e-15cf20c6e085", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03551399428397417, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7eb997b9-50e5-59e0-8a3e-15cf20c6e085"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.05876733176410198, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.057101557962596416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 12288, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06", "validation": {"axes": {"K": 4096, "M": 12288, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.04194791987538338, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/58968e9db4efaafc3ee4bba457398afca703cd0aa049f476f437399dfddf4732.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.05876733176410198, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1ef5e531605f82e019d314bcd63a1457da4f407923461967801d89a06bc92e5b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239366.2158682, "step": 7, "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.057101557962596416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818"}, "variant": "rows_8192"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ff6b72139b04fa9130144a63bab15c6ff519505fd4d6f326cc23ee482a7ec552.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.04194791987538338, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06"}, "variant": "rows_12288"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/164724b8f736981258f69907c85c883cf083dc5f393900970197af72cc45da0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "b195ce43-7a00-5c7e-8570-4081ae872fdb", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.08632527804002166, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b195ce43-7a00-5c7e-8570-4081ae872fdb"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 30, "Nq": 30}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/5182ae66ea455594116b0c41913cb8f9d94a2fedcad7e5a5b4d58b09ca2a88f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 30, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "92594a4b-163e-553b-b0e6-ee5932121fe0", "validation": {"axes": {"B": 1, "Nall": 30, "Nq": 30}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.0024891369976103306, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "92594a4b-163e-553b-b0e6-ee5932121fe0"}}
-{"axes": {"K": 4096, "M": 12288, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea", "validation": {"axes": {"K": 4096, "M": 12288, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.3068883470259607, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 12288, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a", "validation": {"axes": {"K": 12288, "M": 12288, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.08012292999774218, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 12288, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1", "validation": {"axes": {"K": 4096, "M": 12288, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.03814372327178717, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e8cd7bb7a15651b159c4649f4ab20dac14a78ba1ab3ce7494ac9f2b1948d4580.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.3068883470259607, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea"}, "variant": "rows_12288"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/805484a6218332a054439c789defb50bf2195a889cbd859ffab2e6bbe8ed2550.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 0, "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.08012292999774218, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a"}, "variant": "rows_12288"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c8680a8edeb1c05535a2e0c3911596054a84f108b79ebd553df6d9d0878a4e31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.03814372327178717, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1"}, "variant": "rows_12288"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/a5bfe0abdc234ffa77ca9d6fd583bfa6de4e7f6f921a4aa6aef800ac3e17c910.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "9a5607a1-3d45-5614-9c28-556a7235eaf5", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.05036327103152871, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9a5607a1-3d45-5614-9c28-556a7235eaf5"}}
-{"axes": {"K": 4096, "M": 12288, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6", "validation": {"axes": {"K": 4096, "M": 12288, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.07530533708631992, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 12288, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13", "validation": {"axes": {"K": 12288, "M": 12288, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07717035384848714, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13"}, "variant": "rows_12288", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.028228317853063345, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 4096, "M": 4125, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605", "validation": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.01648318814113736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/556b48d2e95940c9e20e93bd8cbcfa92e97c40b958bec2affc7e5c540af69a6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.07530533708631992, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6"}, "variant": "rows_12288"}
+{"axes": {"M": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/515b571e521810128a185851cdedb170a79903d6ad7dcdd4e896540dfc986b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239383.7062342, "step": 7, "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13", "validation": {"axes": {"M": 12288}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.07717035384848714, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13"}, "variant": "rows_12288"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.028228317853063345, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97"}, "variant": "rows_8192"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.01648318814113736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605"}, "variant": "rows_4125"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/133f657a23877ad38fb6e1a13cfe79fca69d3c0b0a42041db5e61af4de8100c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "8ac22de4-c54c-594a-9759-904f1c30f6f7", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.0512183029204607, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8ac22de4-c54c-594a-9759-904f1c30f6f7"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/88d8d74e04b85e837291f3803e67f4fcc0c0f34e2c657e377188c36be21a3217.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "a1487fbc-a4dc-5724-9bd9-0faf4c062b6b", "validation": {"axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.028184604365378618, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a1487fbc-a4dc-5724-9bd9-0faf4c062b6b"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.055546706076711416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05560618685558438, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4125, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda", "validation": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03876070398837328, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 4125, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd", "validation": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.040546406991779804, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030160407070070505, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0ef04f05a00478a89c1b4b0188148efef283f0edd09d456eb27ec1b535c0a063.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.055546706076711416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bca455937434011b259aa80f80861a54352a79daa077a940ed3cdf50ea0dd0d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.05560618685558438, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0"}, "variant": "rows_8192"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03876070398837328, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda"}, "variant": "rows_4125"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 0, "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.040546406991779804, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd"}, "variant": "rows_4125"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7f4080a6c5555bf12e9aaf6757822efc462b9067ead0da2ed669c0e809e545e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030160407070070505, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d"}, "variant": "rows_8192"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/7ed73de575477d876ee2981ece7acd672789ba6fb09b3b83b214de5a92cb7ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "85c2aa46-93b8-5725-ba09-29913e05e1cd", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.051746082957834005, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "85c2aa46-93b8-5725-ba09-29913e05e1cd"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.059363945852965117, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.058005810249596834, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.02641299506649375, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2853a80fc4317a86f06b98b502e7ab616e225fb12947dbfd1a21463a2a8bbac1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.059363945852965117, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/765cfebcce81c1dfa6b40b4e39376d0997ba89bfc006133c7a22d439ecbef9ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239393.0137348, "step": 7, "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.058005810249596834, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b3b99ef0fc63956f514c1d2a3a83e93eeed5b53defdb7dce3c8c7e309b6d6af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.02641299506649375, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557"}, "variant": "rows_8192"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/10717ae953e6012412024c6acb8b0875873bb08ac3b6ffd0754feb989dd839ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "aa86e0ed-fe05-5873-998c-df6ff7963aae", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03146282723173499, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa86e0ed-fe05-5873-998c-df6ff7963aae"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 32, "Nq": 32}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/16b2415f217ef91a9a67466dc67b4f5cf097d249377e14a464187e8006435bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 32, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "b8441206-1742-5b50-a686-e0d86149ca28", "validation": {"axes": {"B": 1, "Nall": 32, "Nq": 32}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.00240721320733428, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b8441206-1742-5b50-a686-e0d86149ca28"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05917341588065028, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.058030262123793364, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4", "validation": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030127210076898336, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/45bc577fdd6156c87de4860bb22460fa8a8f021f385f9baf8b47b572d5d2c3f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05917341588065028, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/49f73531422709582089fb5916de279b062ce9a9cc79d873acc234d20b0f542a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 0, "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.058030262123793364, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/fd1449aad6fde5b85393e2a05b70531c691604ad0ed29800eca0de48a9bd3be7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.030127210076898336, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4"}, "variant": "rows_8192"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/e1530ed27fbd9f4871ae26741dcb8a00be93a96ab263ece6a72bb327ec06e2c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "df7932e0-93d3-5be4-92af-f6eba34890a5", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03204311802983284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "df7932e0-93d3-5be4-92af-f6eba34890a5"}}
-{"axes": {"K": 4096, "M": 8192, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "0882664a-03ac-5037-b167-fa264ffa0639", "validation": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05981408432126045, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0882664a-03ac-5037-b167-fa264ffa0639"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 8192, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe", "validation": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.05757515178993344, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe"}, "variant": "rows_8192", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.08350890874862671, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2587d48e43962811485244f272459d67d7e7ffd396f95c02125b9fd9457aa925.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "0882664a-03ac-5037-b167-fa264ffa0639", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05981408432126045, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0882664a-03ac-5037-b167-fa264ffa0639"}, "variant": "rows_8192"}
+{"axes": {"M": 8192}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df205ade4853b5cd057d8112d25d7a8ea8f3ead9e1b8a431c22457771eedba6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 2, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 583, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 2, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch2", "requested_outputs": 2, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b2-r1/results/images", "--input-save-path", "/task/raw/image-b2-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239400.7528443, "step": 7, "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe", "validation": {"axes": {"M": 8192}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.05757515178993344, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe"}, "variant": "rows_8192"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.08350890874862671, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548"}, "variant": "rows_16384"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/efa0d870b83867f8ea856ba39ba809b330fca342415994531f7ecb7529d799ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "5ad9db13-318b-5717-a1fd-6bafe619bb28", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 17.520813889801502, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5ad9db13-318b-5717-a1fd-6bafe619bb28"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 25, "Nq": 25}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/ec61a985467f7c12f4496197b00f64475ab6ac5f5b4e142527235e7ae823a9d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 25, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "c6f454b8-c97f-582d-a63c-faa18a6f6ad9", "validation": {"axes": {"B": 1, "Nall": 25, "Nq": 25}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.007903106976300478, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6f454b8-c97f-582d-a63c-faa18a6f6ad9"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "6c24a9cf-df14-538e-9876-3431874e3701", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.4345134859904647, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c24a9cf-df14-538e-9876-3431874e3701"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.11248841416090727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.08605304500088096, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/fb4921609679476f4798dd212bf8c2664f5010298a19a7c8b31f339899216163.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "6c24a9cf-df14-538e-9876-3431874e3701", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.4345134859904647, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c24a9cf-df14-538e-9876-3431874e3701"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c9423d86c669bcfe363c735213067aa4e2c7692d87bbbd31b0d5657ef6fb524f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 0, "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.11248841416090727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/112b591be58c8366bf508e157055684de7853bcd091a062d2ad10fa00b45cfad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.08605304500088096, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d"}, "variant": "rows_16384"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/29fa773ea9d4948cbb1d8d03edff83594e9f28d4fdc52ee3186bb05e9b3645b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 25, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "301ea11f-ef1e-5dbf-903c-f53f830b7f49", "validation": {"axes": {"B": 1, "Nall": 4121, "Np": 25, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04081157920882106, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "301ea11f-ef1e-5dbf-903c-f53f830b7f49"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.16125830775126815, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.11023131385445595, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 24576, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f", "validation": {"axes": {"K": 4096, "M": 24576, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.07319077430292964, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/5f7258afce97231dcfe4e69124dffde02e063b9c2a801460c13620e5e1d10b6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.16125830775126815, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/bc16b5eaf4b699260d2bf850ec26ddaa0762588aedf1721e599aaab5cfd4a65d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A red ceramic teapot on a wooden table beside a window, soft afternoon light.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_square_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239509.878295, "step": 7, "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.11023131385445595, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8"}, "variant": "rows_16384"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6d2216c22331e042697018e03e769e31d54511e186e634c64997144158095394.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.07319077430292964, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f"}, "variant": "rows_24576"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/164724b8f736981258f69907c85c883cf083dc5f393900970197af72cc45da0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "62e5d87f-9f4f-52c8-96b4-5c97c8932028", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.10544117726385593, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "62e5d87f-9f4f-52c8-96b4-5c97c8932028"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 30, "Nq": 30}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/5182ae66ea455594116b0c41913cb8f9d94a2fedcad7e5a5b4d58b09ca2a88f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 30, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "5f8f619a-7132-5f12-ac6c-ef19bcb3fbd0", "validation": {"axes": {"B": 1, "Nall": 30, "Nq": 30}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.003906491212546825, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5f8f619a-7132-5f12-ac6c-ef19bcb3fbd0"}}
-{"axes": {"K": 4096, "M": 24576, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b", "validation": {"axes": {"K": 4096, "M": 24576, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.7522421278990805, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 24576, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "89b32f3e-3660-5591-8885-48dffb786922", "validation": {"axes": {"K": 12288, "M": 24576, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.16783020878210664, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "89b32f3e-3660-5591-8885-48dffb786922"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 24576, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7", "validation": {"axes": {"K": 4096, "M": 24576, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.10091953771188855, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/578bf17c5693607e5cbffd8533249b100ee1a1b686c3b750009ab41f6dc8a110.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.7522421278990805, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b"}, "variant": "rows_24576"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/72927b582e12d71d289c2ceccab3f36c6d7c71123b7dabc1407380e5e40d47fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 0, "uuid": "89b32f3e-3660-5591-8885-48dffb786922", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.16783020878210664, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "89b32f3e-3660-5591-8885-48dffb786922"}, "variant": "rows_24576"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/774f7b9a78d8bf988776e979aa8aa5b4c63544a39275df5ee0713d2d6ffdd217.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.10091953771188855, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7"}, "variant": "rows_24576"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/a5bfe0abdc234ffa77ca9d6fd583bfa6de4e7f6f921a4aa6aef800ac3e17c910.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 30, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "5853c35f-a657-5720-8165-27f421255063", "validation": {"axes": {"B": 1, "Nall": 6174, "Np": 30, "Nq": 6144}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06686250399798155, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5853c35f-a657-5720-8165-27f421255063"}}
-{"axes": {"K": 4096, "M": 24576, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405", "validation": {"axes": {"K": 4096, "M": 24576, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.1714622601866722, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 24576, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "3589b4f7-d841-5091-87f0-84fc90925259", "validation": {"axes": {"K": 12288, "M": 24576, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.23707909928634763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3589b4f7-d841-5091-87f0-84fc90925259"}, "variant": "rows_24576", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.07636397518217564, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
-{"axes": {"K": 4096, "M": 4125, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2", "validation": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.032974231988191605, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/b995b47f035c0878afab3db51944cc4a5336bc7f1be525a9b28427542b9d1a32.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.1714622601866722, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405"}, "variant": "rows_24576"}
+{"axes": {"M": 24576}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/7837393a11af2cce97f6b95056b11658016dc12a3f2b7eeed6efa030b5f83cc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A quiet lakeside village at sunrise, reflections in the water, mountains behind the houses, wide landscape photograph.", "response_format": "b64_json", "seed": 42, "size": "1536x1024"}, "request_id": "qwen_image21_landscape_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239569.4210482, "step": 7, "uuid": "3589b4f7-d841-5091-87f0-84fc90925259", "validation": {"axes": {"M": 24576}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.23707909928634763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3589b4f7-d841-5091-87f0-84fc90925259"}, "variant": "rows_24576"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.07636397518217564, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b"}, "variant": "rows_16384"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f3f723c566a4f716349447d7ba8bbbf94118c0c4799385d5bc519ec2b551678d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.032974231988191605, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2"}, "variant": "rows_4125"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/133f657a23877ad38fb6e1a13cfe79fca69d3c0b0a42041db5e61af4de8100c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "985e86ef-7f75-5ae0-a11f-2b16728e0a16", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.07032340299338102, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "985e86ef-7f75-5ae0-a11f-2b16728e0a16"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/88d8d74e04b85e837291f3803e67f4fcc0c0f34e2c657e377188c36be21a3217.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "2f90dd5b-3c00-5a2c-8767-9a38ea6897ac", "validation": {"axes": {"B": 1, "Nall": 4125, "Nq": 4125}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.03707037400454283, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f90dd5b-3c00-5a2c-8767-9a38ea6897ac"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1718691810965538, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.12243430502712727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 4125, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345", "validation": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05945582827553153, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 4125, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a", "validation": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.054225822910666466, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a"}, "variant": "rows_4125", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.07907597534358501, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3d48305d939f178cd8b60aae1986f9fd0004e2649a3b0900ebc4481683cf74bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1718691810965538, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/467ade334037b842da3b6d1aca2336af14fa8934db3493d60e990f77f75a755c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.12243430502712727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7"}, "variant": "rows_16384"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9ac5f664ce8eb9c2a335d65460ff23b7a24ccd094765fd9037d575839ce23fba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.05945582827553153, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345"}, "variant": "rows_4125"}
+{"axes": {"M": 4125}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c42272f17f7d4589e136ad00595a27ae911f07e87dc318b8d20353a5016f2a6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 0, "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a", "validation": {"axes": {"M": 4125}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.054225822910666466, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a"}, "variant": "rows_4125"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/6b26c4b5ae9a8acbdeaace0403da7cd75eae5de4013f849cd77dbd531ef186f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.07907597534358501, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09"}, "variant": "rows_16384"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/7ed73de575477d876ee2981ece7acd672789ba6fb09b3b83b214de5a92cb7ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 8, false], [8, 4104, true], [4104, 4125, false]], "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "881d21a7-fe05-55e3-8a3f-fbed1a05e62a", "validation": {"axes": {"B": 1, "Nall": 8221, "Np": 4125, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06200115615502, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "881d21a7-fe05-55e3-8a3f-fbed1a05e62a"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.11522956704720855, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.12029199302196503, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.07532536098733544, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e494f5abd9f8586b4bd24f473c3cfd45f2ec4b13eadc2e8946cc2a72bd4860db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.11522956704720855, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/3a0b17e7438e60e0a47fd576c2ce85c3d44c26f667f8ee1e0fd4e4f56b8cc5cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "Change the teapot to cobalt blue while preserving the composition and lighting.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_edit_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239587.3535016, "step": 7, "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.12029199302196503, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/bf1eafcbcc197bf05a07281267e6cbf894cdb701ebdb46b14f39eba6596c16ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [0.5], "seconds": 0.07532536098733544, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a"}, "variant": "rows_16384"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/10717ae953e6012412024c6acb8b0875873bb08ac3b6ffd0754feb989dd839ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "f9129b9a-80f1-5c9e-9589-1697a7b4738d", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03661294002085924, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f9129b9a-80f1-5c9e-9589-1697a7b4738d"}}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 32, "Nq": 32}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128_prefix0", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128_prefix0/16b2415f217ef91a9a67466dc67b4f5cf097d249377e14a464187e8006435bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn", "operation": "QwenImage21Attention.local_attn (all prefix segments)", "prefix_segments": [[0, 32, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "1c24ae50-e45d-5786-a2d9-36809250cba9", "validation": {"axes": {"B": 1, "Nall": 32, "Nq": 32}, "definition": "prefix_cached_block_causal_h32_d128_prefix0", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.002510282676666975, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1c24ae50-e45d-5786-a2d9-36809250cba9"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.11258525215089321, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "8344cb3d-646a-5983-955d-aa29071bc411", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.1201779879629612, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8344cb3d-646a-5983-955d-aa29071bc411"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"axes": {"K": 4096, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa", "validation": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.08119806088507175, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k4096"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/aa09b35816f4ec0199d1f339bc9b06750db3dc167f098a51101f6d94b83b1fea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.11258525215089321, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/d57d927e98006f51e8fb72bbbd5ff24d92dadc4be43b5faf0646271bbc8b3a2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 0, "uuid": "8344cb3d-646a-5983-955d-aa29071bc411", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.125], "seconds": 0.1201779879629612, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8344cb3d-646a-5983-955d-aa29071bc411"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/5c64aa0bb4886fb265252031d7f183da548b6dc1d4196cb1e5e2dcccf5134689.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.to_q", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k4096", "device": "cuda", "max_absolute_errors": [1.0], "seconds": 0.08119806088507175, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa"}, "variant": "rows_16384"}
 {"adapter": "Preserve post-normalization/RoPE tensors; materialize the engine's segment visibility.", "axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "capture_steps": [0, 7], "definition": "prefix_cached_block_causal_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/prefix_cached_block_causal_h32_d128/e1530ed27fbd9f4871ae26741dcb8a00be93a96ab263ece6a72bb327ec06e2c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.attn.target_attn", "operation": "QwenImage21Attention.target_attn", "prefix_segments": [[0, 32, false]], "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "c8e32486-c5c6-5f42-bfbe-8d1f7c0a66d5", "validation": {"axes": {"B": 1, "Nall": 4128, "Np": 32, "Nq": 4096}, "definition": "prefix_cached_block_causal_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03464225912466645, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c8e32486-c5c6-5f42-bfbe-8d1f7c0a66d5"}}
-{"axes": {"K": 4096, "M": 16384, "N": 12288}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "3f471772-09e1-5090-a367-35799e617731", "validation": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1156368306837976, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3f471772-09e1-5090-a367-35799e617731"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n12288_k4096"}
-{"axes": {"K": 12288, "M": 16384, "N": 4096}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911", "validation": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "definition": "dense_gemm_bf16_nt", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.10815242305397987, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911"}, "variant": "rows_16384", "captured_definition": "dense_gemm_bf16_nt_n4096_k12288"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.20056561101228, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "c884d64e-9420-501c-b198-7807cdda2443", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 4.497790469322354, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c884d64e-9420-501c-b198-7807cdda2443"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.09004547726362944, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.009442159906029701, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.017903539817780256, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.009230131283402443, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.05460761534050107, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.002057234290987253, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.02799592912197113, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004200323950499296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.00674821762368083, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.006750067695975304, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.05405203020200133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.0019972650334239006, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.02804623544216156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.004096947144716978, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006455938797444105, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.006490407045930624, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 32, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3", "validation": {"axes": {"B": 1, "H": 32, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.04134256672114134, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 16, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8", "validation": {"axes": {"B": 1, "H": 16, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.6162232710048556, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687", "validation": {"axes": {"B": 1, "H": 8, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.23960432969033718, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1", "validation": {"axes": {"B": 1, "H": 8, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.8068100861273706, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 4, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1", "validation": {"axes": {"B": 1, "H": 4, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 10.98438697680831, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04800855414941907, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.001971811056137085, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.01644506072625518, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.004052453674376011, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01087050512433052, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "cacf8279-51df-5a97-a592-9d328a81df13", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.004464325029402971, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cacf8279-51df-5a97-a592-9d328a81df13"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.017077457159757614, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.001979278866201639, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "10493d50-a0f8-527f-8d39-414c377360f2", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.016225908882915974, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "10493d50-a0f8-527f-8d39-414c377360f2"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "2a358cbf-1566-510c-b15e-274202339d18", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004029539879411459, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2a358cbf-1566-510c-b15e-274202339d18"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.004312260076403618, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.004294415004551411, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.014925865922123194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0019850628450512886, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.013812421821057796, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "8471e8e5-d50c-51c2-b946-b92561322441", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [6.103515625e-05], "seconds": 0.004102291073650122, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8471e8e5-d50c-51c2-b946-b92561322441"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31", "validation": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004176128190010786, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [3.0517578125e-05], "seconds": 0.0041802311316132545, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 32, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "503c803e-a524-562c-a980-daf2bad42368", "validation": {"axes": {"B": 1, "H": 32, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.02088306611403823, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "503c803e-a524-562c-a980-daf2bad42368"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 16, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c", "validation": {"axes": {"B": 1, "H": 16, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.06787523115053773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0", "validation": {"axes": {"B": 1, "H": 8, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.11517253704369068, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857", "validation": {"axes": {"B": 1, "H": 8, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.8816024870611727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 4, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de", "validation": {"axes": {"B": 1, "H": 4, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 4.989393078256398, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.054834083653986454, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.0029808501712977886, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.031006361823529005, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.004909697920084, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.009198926389217377, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.008517648093402386, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "a34594b2-2c8d-58cd-b443-630151fec624", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.04977079015225172, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a34594b2-2c8d-58cd-b443-630151fec624"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0024036471731960773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.030632327776402235, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.0048117199912667274, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.008404413238167763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.008421491831541061, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3", "validation": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.04946383694186807, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0023901108652353287, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0", "validation": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.030247369315475225, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1", "validation": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004569757264107466, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13", "validation": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007825418375432491, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0", "validation": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007775096222758293, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 32, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b", "validation": {"axes": {"B": 1, "H": 32, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.039467715192586184, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 16, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff", "validation": {"axes": {"B": 1, "H": 16, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.13199147302657366, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3", "validation": {"axes": {"B": 1, "H": 8, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.23392906924709678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 8, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa", "validation": {"axes": {"B": 1, "H": 8, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.7970105661079288, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "H": 4, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de", "validation": {"axes": {"B": 1, "H": 4, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 12.406822388060391, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.3564747399650514, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 4.4906132887117565, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.1707967766560614, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.015585278160870075, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.030875524040311575, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.011705226730555296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.08746376121416688, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0023189890198409557, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.056300828233361244, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0063310349360108376, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01148474495857954, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.011359761003404856, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.08758193673565984, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0022568400017917156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.055854735895991325, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006105829030275345, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.010980389080941677, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.010989570058882236, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 32, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807", "validation": {"axes": {"B": 2, "H": 32, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.06161692691966891, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 16, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab", "validation": {"axes": {"B": 2, "H": 16, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.6253014388494194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e", "validation": {"axes": {"B": 2, "H": 8, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2609353680163622, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218", "validation": {"axes": {"B": 2, "H": 8, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.9233809900470078, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 4, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc", "validation": {"axes": {"B": 2, "H": 4, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 18.63961925683543, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.09061764599755406, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.0022692279890179634, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03103246120736003, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006090428214520216, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.014651388861238956, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.009273265022784472, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03225407283753157, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0022834837436676025, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.030449194833636284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006085486151278019, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "475965cf-cbad-5baa-b685-479860697719", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006569539196789265, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "475965cf-cbad-5baa-b685-479860697719"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [6.103515625e-05], "seconds": 0.00666813226416707, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.030416897032409906, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0022522597573697567, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.028038349002599716, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006088613998144865, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8", "validation": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.006354016251862049, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006424283143132925, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 32, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016", "validation": {"axes": {"B": 2, "H": 32, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.030310614965856075, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 16, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf", "validation": {"axes": {"B": 2, "H": 16, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.07690511969849467, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e", "validation": {"axes": {"B": 2, "H": 8, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.12507362896576524, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1", "validation": {"axes": {"B": 2, "H": 8, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.9556439155712724, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 4, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44", "validation": {"axes": {"B": 2, "H": 4, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 8.43686280399561, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10575728584080935, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.0038154777139425278, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06242698756977916, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.00763657595962286, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01616377802565694, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.01491745188832283, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10005606804043055, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003065758850425482, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06197610730305314, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007583353202790022, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.014985054731369019, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.014807262923568487, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d", "validation": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.0994175998494029, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.00308423675596714, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b", "validation": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06198076577857137, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "4795bcdd-6f37-5076-9404-a44be707081c", "validation": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007042011711746454, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4795bcdd-6f37-5076-9404-a44be707081c"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e", "validation": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.013635935261845589, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e", "validation": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.013432308100163937, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 32, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "961d1949-75f1-55a6-9950-7c470869589d", "validation": {"axes": {"B": 2, "H": 32, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.05770906386896968, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "961d1949-75f1-55a6-9950-7c470869589d"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 16, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9", "validation": {"axes": {"B": 2, "H": 16, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.1499946415424347, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a", "validation": {"axes": {"B": 2, "H": 8, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2536184713244438, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 8, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf", "validation": {"axes": {"B": 2, "H": 8, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.9129024068824947, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "H": 4, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e", "validation": {"axes": {"B": 2, "H": 4, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 20.85770556004718, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.878054175991565, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 4.5280489986762404, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.41439854772761464, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.03254063706845045, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.07182777114212513, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03552237572148442, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.2588155958801508, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "abb09f82-ee47-528a-905a-f86b658621f5", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0028979526832699776, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "abb09f82-ee47-528a-905a-f86b658621f5"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.14615144301205873, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01586020179092884, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.033811728935688734, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03289396455511451, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.23946930607780814, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.0028513921424746513, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10912138083949685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01061081001535058, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.020939147099852562, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.020583586767315865, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 32, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda", "validation": {"axes": {"B": 4, "H": 32, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.11728813499212265, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 16, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4", "validation": {"axes": {"B": 4, "H": 16, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.7234842558391392, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901", "validation": {"axes": {"B": 4, "H": 8, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.30765239894390106, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67", "validation": {"axes": {"B": 4, "H": 8, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 2.1867749830707908, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 4, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa", "validation": {"axes": {"B": 4, "H": 4, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 35.441113733220845, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "d3373710-c368-59f3-9515-7b961b87d57c", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.17746720975264907, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d3373710-c368-59f3-9515-7b961b87d57c"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.002828259952366352, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.062051201704889536, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01036933995783329, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.02837013592943549, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.011355063412338495, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06526764808222651, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0028808768838644028, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06193388858810067, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.010333027224987745, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.011175353080034256, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.011115758214145899, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.06480937078595161, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.002844560891389847, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06066583702340722, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.010276524815708399, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "645dc728-94d6-5801-a837-9a1f4201b237", "validation": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.010753185953944921, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "645dc728-94d6-5801-a837-9a1f4201b237"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.010692563839256763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 32, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373", "validation": {"axes": {"B": 4, "H": 32, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.05699899513274431, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 16, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5", "validation": {"axes": {"B": 4, "H": 16, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.11679140664637089, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b", "validation": {"axes": {"B": 4, "H": 8, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.14409761037677526, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5", "validation": {"axes": {"B": 4, "H": 8, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.0312290657311678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 4, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "6da01593-e133-55fe-8eb8-69be7217be58", "validation": {"axes": {"B": 4, "H": 4, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 16.0095423348248, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6da01593-e133-55fe-8eb8-69be7217be58"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.2070871451869607, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.008462416008114815, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.10490890173241496, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.015299574937671423, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.030222739558666945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.028602121863514185, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "919246a0-b394-5215-b9c0-afda319ecefa", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.20110497903078794, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "919246a0-b394-5215-b9c0-afda319ecefa"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004404451232403517, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.10427208803594112, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.013297016266733408, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.028794713784009218, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.02881968440487981, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32", "validation": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.2003156109713018, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.004348859190940857, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f", "validation": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10353726986795664, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d128"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392", "validation": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01301031420007348, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66", "validation": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.02794422162696719, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9", "validation": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.026809928938746452, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "noncausal_dense_attention_h32_d64"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 32, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298", "validation": {"axes": {"B": 4, "H": 32, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.11142154317349195, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 16, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1", "validation": {"axes": {"B": 4, "H": 16, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2318596919067204, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6", "validation": {"axes": {"B": 4, "H": 8, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2929818043485284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 8, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5", "validation": {"axes": {"B": 4, "H": 8, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 2.1642743698321283, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5"}
-{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "H": 4, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_d64/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a", "validation": {"axes": {"B": 4, "H": 4, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 39.63709596823901, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math", "captured_definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n12288_k4096", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/52242c387e9b0142be33380829460bac47428c61cd3efeeacced4d883086cd0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.gate_layer", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "3f471772-09e1-5090-a367-35799e617731", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n12288_k4096", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1156368306837976, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3f471772-09e1-5090-a367-35799e617731"}, "variant": "rows_16384"}
+{"axes": {"M": 16384}, "capture_steps": [0, 7], "definition": "dense_gemm_bf16_nt_n4096_k12288", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cdfecb022e906114110dbb7e9699d93328a4f54d2c3ee44cd5285524119c66dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Qwen/Qwen-Image-2.1-Turbo", "model_batch_size": 4, "model_revision": "fc13ba1986fcb905854dc8de2f52eee395f49b5b", "module_path": "transformer_blocks.0.img_mlp.out", "operation": "torch.nn.functional.linear", "process_id": 581, "request": {"guidance_scale": 1.0, "model": "Qwen-Image-2.1", "n": 4, "num_inference_steps": 8, "output_format": "png", "prompt": "A single orange cat sitting, full body, isolated on a transparent background with an alpha channel. No floor or shadow.", "response_format": "b64_json", "seed": 42, "size": "1024x1024"}, "request_id": "qwen_image21_alpha_batch4", "requested_outputs": 4, "sampling_adapter": "Use the unmodified Turbo schedule stored in model_index.json.", "sampling_sigmas": [1.0, 0.978453, 0.95418, 0.926626, 0.89508, 0.845148, 0.704534, 0.414568], "server_command": ["sglang", "serve", "--model-path", "/task/models/qwen-image21", "--model-id", "Qwen-Image-2.1", "--host", "127.0.0.1", "--port", "30104", "--scheduler-port", "30124", "--master-port", "30125", "--performance-mode", "speed", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--output-path", "/task/raw/image-b4-r1/results/images", "--input-save-path", "/task/raw/image-b4-r1/results/inputs"], "source": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo", "started_at": 1792239601.4219952, "step": 7, "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911", "validation": {"axes": {"M": 16384}, "definition": "dense_gemm_bf16_nt_n4096_k12288", "device": "cuda", "max_absolute_errors": [0.25], "seconds": 0.10815242305397987, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911"}, "variant": "rows_16384"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/2ac2a1d1e80676c131af9c5b87b012ab989b4d17087129af1e3efc364a8aee91.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.20056561101228, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "c884d64e-9420-501c-b198-7807cdda2443", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 4.497790469322354, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c884d64e-9420-501c-b198-7807cdda2443"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/793861d1494df006ca325c4276e8aade9eaa15a9dbd82dd3d33d74ffcb236316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.09004547726362944, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/21e5aa860b4bb207de53f96fc4d00b1d318c2b6faaaed12193f620ebbefc69ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.009442159906029701, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bfe6c981523624929104cfa3e2b4d5f4f80000012e0388efdc1efa04224de998.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.017903539817780256, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f432c0d559af1a91ec604bb9dfcd33f18132ad81db7255a6401ffd9d3cb789a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.009230131283402443, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5897e3b872a866c908f982974aa7b80ca7dba00e982f64d0d34802df82562d1b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.05460761534050107, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b4161fe9a82546194c9df7a8fc954778841d917f99a92fae83e6402c6cc4474d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.002057234290987253, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/cced123163ad4947c76601df2d55ae2cb5907a701afecbf4cf44e1ea3324a87c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.02799592912197113, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1b456c5165e08733548e8b07913faf9f402b3c21632c7ebbc2d38239cb354ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004200323950499296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f5e02816c9e6c4bda246bd18b24cf9a7362d3a5e66046222f23fd4fe5eb77a2e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.00674821762368083, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d796ec3d2a2ceb01f649054a35b9cbf9f1184f0a3aaa7d5a3d0cfa8c656399ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 4, "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.006750067695975304, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/982cf3dcd23cf10f12351e3df1be784376463253d56dcea27f9516743a403766.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.05405203020200133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cb5f723948a9906da254f5dbb706a9f2d9000dc1bcabae3d6af421242a63aab3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.0019972650334239006, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9aee725f4c705ab49598a8e1a119fe28ef20d39db9dd86db1a9575de9ac5c567.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.02804623544216156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3b9304c0dabbf07cd0c66290fb61a372920e0d31da3021affc2c1ca1e91e32fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.004096947144716978, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/89edf9a73c3f89eed8873ee550cfd0791251e2a95f6e01541f65cb63a64071bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006455938797444105, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d277927f0658865620cc5e24d1371e47c6bc998d782afd8ad24747b39925ee5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 7, "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e", "validation": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.006490407045930624, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c9a1b59ab2280beae723912c9ff8170322285e1141951c2142668d6eb4790694.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3", "validation": {"axes": {"B": 1, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.04134256672114134, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7c155350a8970f49e963183abf8c263293b42601cea5d8c0a135bf48af359500.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8", "validation": {"axes": {"B": 1, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.6162232710048556, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b03e5f130afa3425eeaddcd5651b0697b399999fa2d6824ca7db4a5026f23d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687", "validation": {"axes": {"B": 1, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.23960432969033718, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/f70a1a8e117981439f190a4c6e55f60399f521bb45bf5a1bae82deee891b4e6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1", "validation": {"axes": {"B": 1, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.8068100861273706, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6d86b779c94225e2ed9381b191d2ff2c71961c3afbb4ba80870d65d4f7bb8d02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241711.9225547, "step": 0, "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1", "validation": {"axes": {"B": 1, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 10.98438697680831, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6307cb0a22b7ac2e920fe9ad9453d89ef42f0b815f7cc695fe30c48ec9ad0395.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04800855414941907, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3f45356cd73e6d1045cf350d684ebac058e60d9b764ff34e85d9a092c6cf32f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.001971811056137085, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ff29237e311cf49dcd7b4bb16aefa60d36f0a0a66155b1519bb66c0ce8edecf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.01644506072625518, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/87a4d3fd3fb1284882109437cd44ae9ecad1dc6ea5a4d2a88923aae98e2fbc87.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.004052453674376011, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a31fb2550e32724de7e7d65dd1c7e867acfe3122a2f4dc882cfa723bac70fdc5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01087050512433052, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f4dafd2827123b850357c26784e25b2907e73366f2560227a238aa0884d620f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "cacf8279-51df-5a97-a592-9d328a81df13", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.004464325029402971, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cacf8279-51df-5a97-a592-9d328a81df13"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0b72eea4024de43033960bf52873fe4af465d06ad4432a952203952ef3675a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.017077457159757614, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/97783ed9bf6d72e13731c39ad639c4cb1d45cf2c2f6112157a0b53ac12fd8bfa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.001979278866201639, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6326bc6250bb1e63da0b5a7e41bf586d09eae6c0da106d3b45b2e325cdace6ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "10493d50-a0f8-527f-8d39-414c377360f2", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.016225908882915974, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "10493d50-a0f8-527f-8d39-414c377360f2"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bc0371619e2c5f0ef19a3b71a8e1dd94c1ad8e5390824f7ca00d4148479eeb87.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "2a358cbf-1566-510c-b15e-274202339d18", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004029539879411459, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2a358cbf-1566-510c-b15e-274202339d18"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2125441614e33de8e5ad2326dca09b45cbeee76b779bd7a1160f4d3a6edd66fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.004312260076403618, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b571a1465fbdad5437b7a98dd5a36fe5ccab95b190660282338669ed7c16b3c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 4, "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.000244140625], "seconds": 0.004294415004551411, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/429b37ccea1cf717aa82f3e5c5799423ba77305a514a6852fbff7f34f18f65b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.014925865922123194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c9fcc64470f06cff7542376da19817b9652a7f1a8f5840bfdc42cc19c023efe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0019850628450512886, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/d83108a7a8aabfb7f213b2bef649506910058d15992121a25fd8aea46c827a2b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.013812421821057796, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a9dae4c8c01ead85bf4cb6c39b36646cde3b749278991a27c56b19fd75f320b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "8471e8e5-d50c-51c2-b946-b92561322441", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [6.103515625e-05], "seconds": 0.004102291073650122, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8471e8e5-d50c-51c2-b946-b92561322441"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/65fd3ad1671d5e277942a3ed73691dba3640e18a4c47f2123563a83f6f5d4883.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31", "validation": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004176128190010786, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/45a46e9b060a9c2765a49b17cd42d8b88a625352c31cb9c9b7486e6130197e48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 7, "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9", "validation": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [3.0517578125e-05], "seconds": 0.0041802311316132545, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/75547a2c62a7b00a958343f9e6d171d057326dd4f776bf04664b8147240913ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "503c803e-a524-562c-a980-daf2bad42368", "validation": {"axes": {"B": 1, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.02088306611403823, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "503c803e-a524-562c-a980-daf2bad42368"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/e405f2f1bb5a266786f48ac34162d6cbbee87032aff59edcd854d4fb16b7539a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c", "validation": {"axes": {"B": 1, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.06787523115053773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/abae80365021b2345eba899010dd61cba48c1e1f7fc2572a9f484c44d76a6002.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0", "validation": {"axes": {"B": 1, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.11517253704369068, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a9ed6a3e5469db99b6fe4f1a95c9e8bb6de1784ba4da869ec3be1a54ed57fb36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857", "validation": {"axes": {"B": 1, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.8816024870611727, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/9963ed274554cd6e176d498a07bf093818db8e4f015d2a5572cec7d92c63a8db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241758.0054827, "step": 0, "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de", "validation": {"axes": {"B": 1, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 4.989393078256398, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e0841729cb92b1cd9d247c730153dca1fdb40c57bfd14950d2df9db6bac60946.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.054834083653986454, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/283148eb29dff8a7dc71d586457543be9f86bcfc0480839c0414d3a52f25c263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.0029808501712977886, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a5a4b72812772ea31d92209ed871796330b88a27474854fc7ba095b6ab224e72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.031006361823529005, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2e6649a91720ca8a911e07694adb93856a6499c0f36cd81d07ddda0ebecb39e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.004909697920084, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bfd39246a5c540352adc8afa6f1f209025ada95bfa033c80a6217f07e8ee61e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.009198926389217377, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/358a8705e9de07aaa9dcd376864dd3dc7195f03bb6bb85d20bd05a171274b58e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.008517648093402386, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/4ed745038d995b1f7b638e5634151b8e63e940bb60d5cb2c9623d8dd122ddd09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "a34594b2-2c8d-58cd-b443-630151fec624", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.04977079015225172, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a34594b2-2c8d-58cd-b443-630151fec624"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/42e18ada4d1fe2c844bb52be33a591cdc7ecadf813a2082a7efe7a8da505b818.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0024036471731960773, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a02b340cc222492791888e9cff78c52833331793121a2c78dcfcbdc6e6b999eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.030632327776402235, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fe742b385cbf48c01ae1753de75ce290d897929df8927ede07b8bcd635b7ca05.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.0048117199912667274, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c56548ccd797ef9eb59f6176c00b2a85175896db33c41a02b347ea639510d460.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.008404413238167763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d57b6686c58237d96bb4a6848142e81cc33ece56196775a5ddf697b206628b4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 4, "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.008421491831541061, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/493a5a17bedd061b42b22f951d095f0e2e1e72896e3e22e279b19d59b90c3950.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.04946383694186807, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f4ec26669e5826b595ccdc08cf9ca1e6038e9acf2f65e51052e6551e760f68c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0023901108652353287, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/deabdb22bb17649c96f781e5cdfd8e7e4ca40d8d7c57de76cd2e0666b9b436b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.030247369315475225, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/01d4ebb18d59fd4e558fcbc47e2996ddd8550e9a92b9313e199ef8f19eb5876e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1", "validation": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.004569757264107466, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5e6ebfd69531b666fd55597233632b55bdd03c58c8da6c80af18d5ec7a9c548e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13", "validation": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007825418375432491, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/47db632032bdbf9839f003e9db86067f1342f3ddc7f8293f19edb48c15129589.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 7, "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0", "validation": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007775096222758293, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6c4021dc447efc2f8b874bf19d7c29b7c6cb9c84bf5a91c34183ed042d9a6cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b", "validation": {"axes": {"B": 1, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.039467715192586184, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b6aa794d0622e0d7c9d7ee8dcb39c463ef97770d97e1595d55dfb1357b50e326.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff", "validation": {"axes": {"B": 1, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.13199147302657366, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cdb619e744bcec546fff6de2bac95d97ce850b0aff47ddc3a9ae580ff8695dc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3", "validation": {"axes": {"B": 1, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.23392906924709678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0546fe692e1c0d4bff54383bca8f443011a4281afd7727698e5ced6e0def8653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa", "validation": {"axes": {"B": 1, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.7970105661079288, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 1, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 1, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3af97dce0be5efa434edbee42cfb9793ce2e9d56410836623eafaa5f6355ed60.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 1, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video", "requested_outputs": 1, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b1-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b1-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241772.0388129, "step": 0, "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de", "validation": {"axes": {"B": 1, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 12.406822388060391, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ba010bc4d143583f885320420cda12789bc2e572d95b0a28906c7222ae587e65.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.3564747399650514, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 4.4906132887117565, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/8aab9a9ed5a55d42f08e4d28d07e43177c927d5bd53db3b1b677154c6005af12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.1707967766560614, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/aa7c481b31ca35f3ae086bed2d8f7a4369ae6499749e53b3aa42217a73065653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.015585278160870075, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/097602210ffcc0e9d882f3b7706633a2d97decb565d197c5a1132329f734bab9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.030875524040311575, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7573886bd67a7f544f14b46babc98f2339aa19f633be64c09dff0e9338f54fe6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.011705226730555296, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f6c4bc7ef6f8cc8bbda7b0626b911f4e3ca2122928f67bb46312f8517c544b0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.08746376121416688, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/12dc05d41935be6b0e5597c78ca18e9a1f365c99ba7c5a1f70c4033050c2d3c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0023189890198409557, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6381269f3a80c566184b5cbee5460bdaf10e31bad9bf06cefe695978c98aeeee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.056300828233361244, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6d53eff0bac33b16776ea563afd8fcad726e3fa7b6f27b05c7a0b2cb968eff06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0063310349360108376, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/febcaa41d7b8d26203636b176e2c36818718b6005269f6ba962646337a105497.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01148474495857954, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d7a7b40eb46f6a75bfb7add8dc3cf774115c4a98f1183361dedaea6fb5571fd0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 4, "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.011359761003404856, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/55d34f8ed230e27f4420fa9499b9f182a87c0dcbb359ebed2c572d24d59267cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.08758193673565984, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7c401218dc23f058af594381bf2ce51db4a1ae46d2fc3be5588cdfec3c58e0eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0022568400017917156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a5f55d6f057b45ab8e6cfbf3349834a86beed72b38e27fc01db1c1ca28e205e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.055854735895991325, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e35850b82a4fde3f2c9f2b44030dc409695d60887543fa082adb2dfa0c55075e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006105829030275345, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/57c7b2b1eb91ba22fd2a6efd669eb663c68d4a42a9effa03f6040e469ea9332f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.010980389080941677, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4f0dacfd1e56a365d4c30cf0f6e9c92bf3eee0c11c12e8a3d185cb164cfacbb5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 7, "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c", "validation": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.010989570058882236, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/fe3fb7acba9665f059ceaf13ed5e823d9c90686316279df7a6bf07f795039cd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807", "validation": {"axes": {"B": 2, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.06161692691966891, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/98c5b9560f3d9a8ba3133adf125a9d11be01a47b739660f4940c27987d7b2cb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab", "validation": {"axes": {"B": 2, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.6253014388494194, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8d0b7c04842036e5b642f5c9e746897d9de74bb3eda689858beb4e3f908e1dfb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e", "validation": {"axes": {"B": 2, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2609353680163622, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b7a1d71546e723e1c9a9969f524109f46005fc85d6b667d6e0e83d00e0630e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218", "validation": {"axes": {"B": 2, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.9233809900470078, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b0b59766fd7c1781275bb7e5029da80ad4bfaf51b9c21e8452e19b36bd9b9d0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792241953.9250946, "step": 0, "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc", "validation": {"axes": {"B": 2, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 18.63961925683543, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/adec87acc9de1d195c2646cb04593537ef8a01365768d2394eafe16dd999e991.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.09061764599755406, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3010017dac0ec793a8e7587561d111ab067160e2d1ef921edaa5424f582b7e16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.0022692279890179634, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/52b2132db1569942dd9adedca02a096ee6100ba701a3c1bc82fbc0fc0b9d4414.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03103246120736003, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5e7d4c5c27a95acc9a33ea4d8532a641856b2325565a9caaf3e3a9b3f6f189cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006090428214520216, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bd87b065b49c346452a46e3b989ccad0251500682c241dff27355ff17be9ffc8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.014651388861238956, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4a340cfd7e2e349ade624205639aec26f9f10f9a2ddb404e55fe0c72cfbb7bc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.009273265022784472, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/d0466745315073f3b900499277298e8c4df7d035a7291d91365a8b7d57681f60.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03225407283753157, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4f09df34bbb4148be0197189046408ea112b188af42fad11845cb4a9354b6c51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0022834837436676025, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/624f3b4068ccadfc3e81d536c052bd10aa3d4eedf2184474f425bdcc92225f9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.030449194833636284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/03b3dbdfe994e8e6caeabc055ea6872581cc5ef594c7c7bd3a81291d7929d70b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006085486151278019, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/41b97cd6a0b571e179b13780c968a80425800222341f2b4cfd7c0278b3ccc05c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "475965cf-cbad-5baa-b685-479860697719", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.006569539196789265, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "475965cf-cbad-5baa-b685-479860697719"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/774bbbf14aa4bfd894e5c2f09f3a4fb555bdecfbf57bc7ec8c152a9749f114f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 4, "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [6.103515625e-05], "seconds": 0.00666813226416707, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/711bed213b009a600243d4f530e2d10a817da768e604f5b0204f1866c5f37f1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.030416897032409906, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5018049bb7e1a47e6d05a84befd96df1bea51b8357a3d2ba52ef9c9b41c6b186.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0022522597573697567, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6282ff2b4e962b8ed0b5efd725fb77a70afc9c2f04daf7fc7ddf59359da05c56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.028038349002599716, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e413095fe96f6bd2e4d23a59c75999ff031b7c391c7f1f3ce6afd72ddcc09a2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006088613998144865, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c0c4f8305404571a75783a3bcd1befb00f7edaefdf84296ab6f10d0386e3d65a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8", "validation": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.006354016251862049, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9abb0dad74d23a188f93f3c017513dc4ab050c0dbebf5a01637ba5fb61601556.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 7, "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668", "validation": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.006424283143132925, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/a82e4ffb152558a9bf267419597ab378797101958d6bf7530ba1f25b6686a2b9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016", "validation": {"axes": {"B": 2, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.030310614965856075, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/86e79d29e9f09952cf348b7ed7a767e7be06c9b63d12bae34a1f6d5b510ac196.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf", "validation": {"axes": {"B": 2, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.07690511969849467, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/97e2496f90361ab146e74eaf5884427a79aedd5b705cbb34a5361d3baf42aeb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e", "validation": {"axes": {"B": 2, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.12507362896576524, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/bd4774e3d42e8794030001de6ba2c474d0ea76bac76a973e3ecb36fb2a37a3e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1", "validation": {"axes": {"B": 2, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.9556439155712724, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/fc0616dacda9dca25e9fdaba0d6bbf49bc9849fef05d748db5e65bf793fb2d98.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242006.0334532, "step": 0, "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44", "validation": {"axes": {"B": 2, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 8.43686280399561, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/fd7c3482e9774d41d70a79cebfbb253b823b857809a39f22bdb6101db8f97aa3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10575728584080935, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c446eefba96bda4d06adba9c9347bf0ae90d330135033d1cdbb1c02ca03b9c5e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.0038154777139425278, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/dbf89936000795742b734c37465068c640828da8e8f8b23fa28e429dd07e6eba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06242698756977916, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f460603b964888235d48514813718de8ecb0b9e02b9849767033cf9c9b1be285.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.00763657595962286, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/393df4e7dc3bf41de7a9a06cc096c3eee5592816e29a38d8e1e66930d8f15eea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01616377802565694, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0cd13e0ddadf9f0510a5cd7115e948fa521a1858e817840479221aa45f0e1f8d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.01491745188832283, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/65a0cdfa2530d0f36fb27016331d1cdb1615d6d3fa2d70a86bd8201d2682f2ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10005606804043055, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cc41afb741391a57f878fddf7a0699bb25bc923e8a3de79fca126859c5205583.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003065758850425482, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0b2803a8cb4a1390c3e3ad4a503413c32981793a1a7a7f000fb3687e8d5bb4e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.06197610730305314, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c3f25e175f8c5dfb3f7eb2c4fd833aa02233c9073cd5a76160bbfa9a203d5e7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007583353202790022, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1a89097b8367073fbb8343344b969f961fce6f7a58bd99da28935f3fb1ebde72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.014985054731369019, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1926d870bb9ed06fa070059ee3d66824f667ca72baf513b715ff0350248fbf12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 4, "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.014807262923568487, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5e1577f1b2aee0f9321d34135db5adec510e70c4ff08787b6d6aca8827a20800.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.0994175998494029, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7847fa735a8f8c5156dc88dc9526883dd7d2320cfe8679076557f4c30b4359e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.00308423675596714, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9f3460de3afd3264471b55ad9218e76a58e1bd79366d9bcf89689616f80e3293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06198076577857137, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d649fa3970e8ddac8022e57ade6c44336a37ac555d1981a0c16ca0aee443ea11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "4795bcdd-6f37-5076-9404-a44be707081c", "validation": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007042011711746454, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4795bcdd-6f37-5076-9404-a44be707081c"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/eb36e9422c7cf86d220f49952941a77e94c0646e76a5a65fd4e6674eb6ec7904.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e", "validation": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.013635935261845589, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8fe56f861959b51196e10bb9b2ae441ac4486fe13724b7812b9dafcb37ad62b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 7, "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e", "validation": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.013432308100163937, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/c32d234743cf2934ac3bd116ee7854784e4fbe14abcd4a14b3a03578e2d52226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "961d1949-75f1-55a6-9950-7c470869589d", "validation": {"axes": {"B": 2, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.05770906386896968, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "961d1949-75f1-55a6-9950-7c470869589d"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/2cace15183b9560b5c3a501b3d67c5f8a8d00e5f581464489250dd92933994f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9", "validation": {"axes": {"B": 2, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.1499946415424347, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f0709e3829436a59369b4e32f6ce8eaa3622408228e5e06dfcb73dab74f5577.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a", "validation": {"axes": {"B": 2, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2536184713244438, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/66c3a3532532ecd91df77e080a1a0ff9a3f4a5966f54158aa8fd3394bbf9a75d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf", "validation": {"axes": {"B": 2, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.9129024068824947, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 2, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 2, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a0e701b7b037fce128e865179f573ae72cb7512b3876a8e598a7cc3a19d324f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 2, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch2", "requested_outputs": 2, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b2-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b2-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242022.0743153, "step": 0, "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e", "validation": {"axes": {"B": 2, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 20.85770556004718, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c31ed1dd13f35216d39e35f535d31953c2e181ab80b42713acc300d2c3d381ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.878054175991565, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 4.5280489986762404, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/37f0ba5b5c1b52e5a97ab7fe3fdcdb561f445d68c22e8f37dc5e7a95ae8f95b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.41439854772761464, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/364956f348a9133d9142f0255cd16c865a466d06647e5ca28e5d77d78bb8436e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.03254063706845045, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f2556bf7ca169063e0a2020c7553acae369f1fa72fdd118c91942ab944a6bd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.07182777114212513, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4778b0dea33cad3ceed92ea273af952e12faea6ae601411fb45f9b565e0597e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03552237572148442, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/635137b3d81cc4dd83db77d34a0a1e41325758e70f42a664078b883ffaca91ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.2588155958801508, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ae747a2ec6853194b019c473eff7fc8d817633dbf01521a280f93371be947dfd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "abb09f82-ee47-528a-905a-f86b658621f5", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.0028979526832699776, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "abb09f82-ee47-528a-905a-f86b658621f5"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a31816421975b3858ba9bade08349aae2601b33b97cc1bd98949233a24e74492.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.14615144301205873, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b679d2b17201bb4726e587134eead62f8d43eabdb898fb8590350e83a64b644f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.01586020179092884, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fe488dda5f1935cb24923ba16a84a8a2bfd8e09e643d060241852cbe16b343c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.033811728935688734, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/336d298af4065f77b2745ad627b16cecfc210961e7cd7f03e7811329fc46e985.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 4, "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.03289396455511451, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/be4426ac7c7f9d97045c85b946f0f0db830b0bfebe7e99e64e695fbdbdc775cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.23946930607780814, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8a570e96f2f6df0c84fac6d82c51255a168059ba49db1b45b340814170d0a78a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.0028513921424746513, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/66e8f5b1ce60519a1ae900eb57ebd72b71570d435c94613c891887437b8b4f6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10912138083949685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/36f7fdef855507afa6a5961da446766b60711fa0692e01537ce2f1399341f027.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01061081001535058, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 2352}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/950be58dd29df63051503b9927fc8f85a8f22aac673e06098559146c44bba348.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.020939147099852562, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2352, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a128004894205e6e95fe9589418209885a17a4e8c4cd31e3c374edb44fd1352b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 7, "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3", "validation": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.020583586767315865, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 9, "X": 24, "Y": 14}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/076a2eff76e16fa504d9f089868859adc02144bbac05eb521cbd17fe55174854.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda", "validation": {"axes": {"B": 4, "T": 9, "X": 24, "Y": 14}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.11728813499212265, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda"}, "variant": "grid_9x14x24", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 9, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6e0a1c765dff714d2fc84ac85797f5885509c16b48b1d1e054aa9f456164d1df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4", "validation": {"axes": {"B": 4, "T": 9, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.7234842558391392, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4"}, "variant": "grid_9x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 17, "X": 48, "Y": 28}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6611bff57cf8e17b278e8a3d1312e8e1b35d316dd54c42fffbe05325d775133e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901", "validation": {"axes": {"B": 4, "T": 17, "X": 48, "Y": 28}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.30765239894390106, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901"}, "variant": "grid_17x28x48", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 33, "X": 96, "Y": 56}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e082ee0b4559b74142ece47d8528fed46c8df3761c0f3f1ef2b37d6a87de9b48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67", "validation": {"axes": {"B": 4, "T": 33, "X": 96, "Y": 56}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 2.1867749830707908, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67"}, "variant": "grid_33x56x96", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 49, "X": 192, "Y": 112}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ea644641dafb04a96f9677985426305371b04a45999ec0d938d5eb5193c350dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A red ceramic teapot on a wooden table beside a window. The camera slowly moves closer as steam rises from the spout. Soft afternoon light and gentle room ambience.", "seed": 42, "size": "768x448"}, "request_id": "ltx25_text_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242247.1026254, "step": 0, "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa", "validation": {"axes": {"B": 4, "T": 49, "X": 192, "Y": 112}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 35.441113733220845, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa"}, "variant": "grid_49x112x192", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/73659cd428a2c0d0fb25ae46514c9697d1ac106f6c76e6d411054e060ac19c72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "d3373710-c368-59f3-9515-7b961b87d57c", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.17746720975264907, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d3373710-c368-59f3-9515-7b961b87d57c"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/adf8b958756b6c5456395607dc101bbf4babcc392c8dddf0daaf5e7988fdc0dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.002828259952366352, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ac9cfc44c2fc20a87353a40849946e0d5a39dcfd0d04fa74370c2658532ea341.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.062051201704889536, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e1012492f0b34016af3e1acc33f91e691f25bacb5909e985b1e61f5288922b1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01036933995783329, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/34fd06382976c9e5ea5abbfa690437b3e6f0a00cb34b1619806ab7e8d208dd59.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.02837013592943549, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0bc98e2efccf97fdf3a051c8d2c7fae2880a72374422c82d65d5f14a5ff1eea2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.011355063412338495, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/81e125e358f1b6d1bde96d3fe924fffe68bdadfa8dcccbde6573dea653260b0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.06526764808222651, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/be694ffea3a4793cc89856d6ffaebe59c32ca2d595c3a38ac2e74ee3315894c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0028808768838644028, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/44165e2052f3530b8661169e52cac279c9f4f97a07434baba270b19941bdfed6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06193388858810067, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/dc3c34114e51602b6b757e556412e1c7f2eb629258889edde6b2e93f6d6d0e0f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.010333027224987745, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7fb91da18ba2b30f4fdd0dd7d8993549d98e86f7671ce9c5e36a5dd3ec22d910.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.011175353080034256, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6ed79d862da987f131ea8a0b6cf38c3d784a77baac69e2838e30eb95753559ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 4, "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.011115758214145899, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/1dc486c0c8c615fa83607e7d74c6ce1c502d7c9059cc3a9d73ff93dd76936d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.06480937078595161, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/47887c9822e410c699f8ccbb75e31b76215477a9792609d9cfca9f2b98c9b323.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.002844560891389847, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/bb81a841233c5f47be77ce44b26aa95bd0786b68874539cd78e3e16190b248b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.06066583702340722, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/27f9e4d855cc6eb836dba6d93c966cdfc3843f309ec845aad0484d684bb100dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.010276524815708399, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 51, "Nq": 1120}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1f0e5f2116b8b9bf8def3832fe6f10d9acab52d71dbc59864a45d7f63cb26ba7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "645dc728-94d6-5801-a837-9a1f4201b237", "validation": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.010753185953944921, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "645dc728-94d6-5801-a837-9a1f4201b237"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1120, "Nq": 51}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/abc7cb1fb3f8f8c4595959440175a159e57211d179c005865e285de050a398aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 7, "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120", "validation": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.010692563839256763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 9, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/058d7df956e2835de0d256063d5e140b40ee61d7454572db692f5b8203e316c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373", "validation": {"axes": {"B": 4, "T": 9, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.05699899513274431, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373"}, "variant": "grid_9x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 9, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/fccb4672737c62149a905287677b6ca1f8cb82e5c8683e21e4d863e50fbe3929.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5", "validation": {"axes": {"B": 4, "T": 9, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.11679140664637089, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5"}, "variant": "grid_9x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 17, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/27dd5e1537155a92dbce19158a59804f1a3808e19d7d8730cc3fb15573361683.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b", "validation": {"axes": {"B": 4, "T": 17, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.14409761037677526, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b"}, "variant": "grid_17x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 33, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/46aec19ef278cd6f62346d49185e694f8b1b21ea80d43fc9961ee799ae4ea9ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5", "validation": {"axes": {"B": 4, "T": 33, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 1.0312290657311678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5"}, "variant": "grid_33x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 49, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7413f14c268624ac29d9d00451d1d41936027ecdaf67bb7ec5084aaf51199108.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "reference_image_sha256": "e6f2095f119b4ca59cace7b8969cac987408fb38d80a76e68ffff043a71ecc80", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "input_reference": "/task/raw/image-b1-r1/results/square.png", "num_frames": 49, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "The camera slowly moves around the ceramic teapot. Soft steam rises from its spout while the table and window remain still. Quiet indoor ambience.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_image_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242361.3072696, "step": 0, "uuid": "6da01593-e133-55fe-8eb8-69be7217be58", "validation": {"axes": {"B": 4, "T": 49, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 16.0095423348248, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6da01593-e133-55fe-8eb8-69be7217be58"}, "variant": "grid_49x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a590229d40d8febdc16a1eaa393696a62fe4fa658ec4dc5a94c46d5a18b0ab46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.2070871451869607, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fecadb73a7d45ca1ad72ae112a165ff81c539bed4b65754bd2e70b3fae36792a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.008462416008114815, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c64f2f27fbfdb4e9633ba25210464ca7caed3aafa02a5d722aca732e432bf272.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.10490890173241496, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/968f86a7091fb914372095ad6b33d130fb2987022c3f2c93c172a8c25c9a5653.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.015299574937671423, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6d47140ef82be5b88da342531918c9c66723d6423fb17ed2280ccf8ccc69a7ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.030222739558666945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3839b8c636327b9999874fb4fa41e0dfd47e64a8ffb9ae2b63d0a09d89b7deb8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.028602121863514185, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e3cfeafe29ba4ccc424673195bff1e85a8605d55d755c2aa94459fbfd2c9d53d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "919246a0-b394-5215-b9c0-afda319ecefa", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.20110497903078794, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "919246a0-b394-5215-b9c0-afda319ecefa"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7dfffb29789c57c5c833f443bcaaaccad9577b37a83f9768de4811f5fe088b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004404451232403517, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e357b3d87f94327c5b844eb98c497525c6e86c0d5b5eab9f7cb28987f16aa86a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.10427208803594112, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/359b3cd62377303671ce84039f6fbeb1fee2a37111d95d4ca0f9556c6b6f2ad6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.013297016266733408, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/56bb13dd44f373a0b5d110ec76a266f9db188dae26525813f37193ab71828881.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.028794713784009218, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a0e918cdaa198fe06cac9bd2c8ef76648333620ec31b75131d43497efe6c5f88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 4, "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.02881968440487981, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/4ecff6d01b884e4bde1f794acee60c5ccc8c1bac3500513e791ade113f655a67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0625], "seconds": 0.2003156109713018, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32"}, "variant": "attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b8a468093af919ab0a59b5b41c87d3d8dffd8370b3bd636a498b5f6eee148f4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn1", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.004348859190940857, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202"}, "variant": "audio_attn1", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d128", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b2f79fd785485bc4a70627c7440b6b61af74f412d24208756f12511d45787b2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d128", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.10353726986795664, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f"}, "variant": "attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 1024, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f1c47c5260c2a624ab4d764d45c4d3fb29499b1d588b1af1d2004eed4da59ea9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_attn2", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392", "validation": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.01301031420007348, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392"}, "variant": "audio_attn2", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 126, "Nq": 2560}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b23182a122264c0f9362b803e9a72c25f11d1bd4f8ab957c2451fe3a278ac6cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.audio_to_video_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66", "validation": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.02794422162696719, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66"}, "variant": "audio_to_video_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "Nk": 2560, "Nq": 126}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "noncausal_dense_attention_h32_d64", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1b40031b1d965be1c6ec6b76e031e214bcf845fe43936dc04ed17e422590d0db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "transformer_blocks.0.video_to_audio_attn", "operation": "LTX2Attention.attn", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "sdpa_backend": "math", "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 7, "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9", "validation": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "definition": "noncausal_dense_attention_h32_d64", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.026809928938746452, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9"}, "variant": "video_to_audio_attn", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 18, "X": 16, "Y": 10}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/462e74c323d7587b91648e116007e772d5fbd962280981391b58e1b9989dfb9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h32_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298", "validation": {"axes": {"B": 4, "T": 18, "X": 16, "Y": 10}, "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.11142154317349195, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298"}, "variant": "grid_18x10x16", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 18, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/af373e4f2d1695d315be0b58e993105ee1624be85e6a284e893fd116b9e9e8df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h16_3x7x7", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1", "validation": {"axes": {"B": 4, "T": 18, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2318596919067204, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1"}, "variant": "grid_18x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 35, "X": 32, "Y": 20}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/205af6e016173fee56a83608a06e25d1799905a1e18a1f000946968143f1431b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6", "validation": {"axes": {"B": 4, "T": 35, "X": 32, "Y": 20}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.2929818043485284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6"}, "variant": "grid_35x20x32", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 69, "X": 64, "Y": 40}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/556acca09dd870fc2ff8429c35d61ae784b0fe36ae61c7fbe2acb1381971393a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h8_3x5x5", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5", "validation": {"axes": {"B": 4, "T": 69, "X": 64, "Y": 40}, "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 2.1642743698321283, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5"}, "variant": "grid_69x40x64", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
+{"actual_attention_batch": 4, "adapter": "Capture the complete grid after normalization, query scaling and RoPE, before output projection.", "audio_precision_adapter": "Select PyTorch math SDPA for every dimension-64 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "audio_sdpa_backend": "math", "axes": {"B": 4, "T": 121, "X": 128, "Y": 80}, "capture_steps": [0, 4, 7], "conversion": {"conv_vae_distribution": "comfyicu/LTX-2.5", "conv_vae_revision": "1026e00f808dbd9601911479b22d65b760737dc8", "converter_revision": "efe46a5b177af67c64b3666da8ddddac7ee12163", "original_model": "Lightricks/LTX-2.5", "original_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "public_distribution": "pinecoresystems/LTX-2.5-serving", "revision": "93deb76408feee57d04b3cfeed3ded56a7a068ee", "source_sha256": {"diffusion_models/ltx-2.5-22b-distilled-transformer-bf16.safetensors": "31eb3cad89b9e54e99dd3baf286f70825ac4f6c660a70d9184d895be76d7bff4", "text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors": "ef7243612fdae7a75cb4d5cee9433e81380675fb6c213bd98ae74a9cd16561d1", "vae/ltx-2.5-audio-vae-bf16.safetensors": "c52733d37f6a7fb7949c3dc0fb468c6cb2169e4d836983a73babb9f0d54837a5", "vae/ltx-2.5-video-vae-bf16.safetensors": "847e14ca7f3355debca0cea4eaa24ac0fbcdf0061da054ac89ca638a869ddba3", "vae/ltx-2.5-video-vae-conv-bf16.safetensors": "685b06ee3d9b2039647698fc4ea33175112462fc374e2777312c907897dfce8d"}, "text_encoder_adapter": "Lossless parameter-name mapping, including vision/audio embeddings and tied LM head.", "validation": "Every component loaded with strict=True; all source file SHA256 values verified."}, "decoder_adapter": "Retain the original optional keyframe tag; use the complete ordinary-video decoder path.", "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d63cf29cf8ca873c1cc6990699c491d629a234cbffb27a612f984490395300e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "model": "Lightricks/LTX-2.5", "model_revision": "49b093a39c19169a8a81721dcc541e07e945232f", "module_path": "diffusion_decoder.attention_h4_11x11x11", "operation": "LTX2VideoVaeNeighborhoodAttention", "request": {"enhance_prompt": false, "extra_body": {"use_diffusion_decoder": true}, "fps": 24, "guidance_scale": 1.0, "num_frames": 121, "num_inference_steps": 8, "num_outputs_per_prompt": 4, "prompt": "A small wooden boat crosses a calm lake at sunrise. The camera follows smoothly as ripples spread behind the boat. Birds call softly from the shore.", "seed": 42, "size": "512x320"}, "request_id": "ltx25_long_video_batch4", "requested_outputs": 4, "sampling_sigmas": [1, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875], "server_command": ["sglang", "serve", "--model-path", "/task/models/ltx25", "--model-id", "LTX-2.5-Diffusers", "--host", "127.0.0.1", "--port", "30400", "--performance-mode", "speed", "--scheduler-port", "30420", "--master-port", "30421", "--attention-backend", "torch_sdpa", "--enable-torch-compile", "false", "--enable-breakable-cuda-graph", "false", "--warmup-mode", "off", "--load-diffusion-decoder", "true", "--vae-tiling", "false", "--vae-sp", "false", "--input-save-path", "/task/raw/ltx-b4-r4/results/inputs", "--diffusion-decoder-tiling", "false", "--output-path", "/task/raw/ltx-b4-r4/results/videos"], "source": "https://github.com/sgl-project/sglang/tree/684fea8a/python/sglang/multimodal_gen", "started_at": 1792242401.3913639, "step": 0, "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a", "validation": {"axes": {"B": 4, "T": 121, "X": 128, "Y": 80}, "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 39.63709596823901, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a"}, "variant": "grid_121x80x128", "video_precision_adapter": "Select PyTorch math SDPA for every dimension-128 transformer attention call; keep BF16 inputs/outputs and FP32 intermediates; normalize output storage to the engine's sequence-major flattening contract.", "video_sdpa_backend": "math"}
 {"actual_requested_output_tokens": 107, "adapter": "Retain all live pages byte-for-byte and remap their page IDs; omit unreachable allocator pages.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 16, 1, 128], "axes": {"B": 1, "MP": 2, "NP": 2, "P": 16}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "engine": "SGLang", "engine_block_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/f85737bdc0d96ac0bc073f36bdc22f7a459f9531c81e826431d9d26cc9974ef9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/f85737bdc0d96ac0bc073f36bdc22f7a459f9531c81e826431d9d26cc9974ef9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "03674a4fc8fea34b383a3015f41b09b0c441f14bdbc25e218c1305307d6f2f08", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QSAIndexer.select_decode_tokens", "original_physical_page_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/1cef83d61ca56b0bacc92e5c55626ed3010575136bc32e54db7db753ea383488.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "output_adapter": "Serialize the engine's selected set in increasing logical-ID order; preserve membership and -1 suffix padding.", "padded_query": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/d52bde5a8d11428bb129ece05fc0404b150685d4ecad5d2cc1d6d9efb0779802.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "query_adapter": "Remove verified zero query heads inserted for the backend's eight-head tile.", "request_id": "qwen_next_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "691b7d0c-2cf5-57f0-8008-6c37d24d2ced", "validation": {"axes": {"B": 1, "MP": 2, "NP": 2, "P": 16}, "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "device": "cuda", "max_absolute_errors": [0], "seconds": 0.49019037326797843, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "691b7d0c-2cf5-57f0-8008-6c37d24d2ced"}}
 {"actual_requested_output_tokens": 107, "adapter": "Snapshot after the current-token KV write; retain every live page and remap physical IDs.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 64, 2, 256], "axes": {"B": 1, "MP": 3, "NP": 3, "P": 64}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/f59183baa1c3a496bfa8fc74042c98152632cd5f5b00f351494abd530499419c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "03674a4fc8fea34b383a3015f41b09b0c441f14bdbc25e218c1305307d6f2f08", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QwenSparseAttnBackend.forward_decode", "original_physical_page_ids": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/4d8c659ebc7c8ea311eb7e327d2047320b4e614bdbf74b067ee4fada5a550a30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "qwen_next_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d15d9f9b-cd28-5c5f-aea1-a0b1c33d63e6", "validation": {"axes": {"B": 1, "MP": 3, "NP": 3, "P": 64}, "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.12573457602411509, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d15d9f9b-cd28-5c5f-aea1-a0b1c33d63e6"}}
 {"actual_requested_output_tokens": 107, "adapter": "Retain all live pages byte-for-byte and remap their page IDs; omit unreachable allocator pages.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 16, 1, 128], "axes": {"B": 1, "MP": 2, "NP": 2, "P": 16}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "engine": "SGLang", "engine_block_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/f85737bdc0d96ac0bc073f36bdc22f7a459f9531c81e826431d9d26cc9974ef9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/f85737bdc0d96ac0bc073f36bdc22f7a459f9531c81e826431d9d26cc9974ef9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "03674a4fc8fea34b383a3015f41b09b0c441f14bdbc25e218c1305307d6f2f08", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QSAIndexer.select_decode_tokens", "original_physical_page_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/1cef83d61ca56b0bacc92e5c55626ed3010575136bc32e54db7db753ea383488.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "output_adapter": "Serialize the engine's selected set in increasing logical-ID order; preserve membership and -1 suffix padding.", "padded_query": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/5cc2958486887b349321f89b318f701f38ac21ed7ceae16a12e92f691d904a3c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "query_adapter": "Remove verified zero query heads inserted for the backend's eight-head tile.", "request_id": "qwen_next_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "43f45cd3-694c-5186-a45e-16e2afec402a", "validation": {"axes": {"B": 1, "MP": 2, "NP": 2, "P": 16}, "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "device": "cuda", "max_absolute_errors": [0], "seconds": 0.0016061612404882908, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "43f45cd3-694c-5186-a45e-16e2afec402a"}}
@@ -458,231 +458,231 @@
 {"actual_requested_output_tokens": 7948, "adapter": "Snapshot after the current-token KV write; retain every live page and remap physical IDs.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 204608, "label": "256k_long_output", "locator": "requests[74]", "model": "claude-fable-5", "output_tokens": 7948, "reusable_prefix_tokens": 204224, "reusable_prefix_tokens_256k": 204224, "role": "main", "selection": {"field": "output_tokens", "note": "Long decode trajectory within the Qwen model context limit.", "percentile": 99, "target": 7948}, "session_id": "1b577bfa02cd2c421c0b59c5cac702b3c287", "t": 3693.72, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 64, 2, 256], "axes": {"B": 1, "MP": 3260, "NP": 3260, "P": 64}, "cache_predecessor_locator": "requests[72]", "capture_decode_steps": [0, 1, 2, 3, 3973, 7946], "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/fe9ec2d951399c86bf0eeaefcf725fb5a0fce6eb1067e21b7a81d910b94adcbc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "93a0fe2aa6d6a9252367be8da47bf2577fed128d48c71bd4d3371f3c8f22aa25", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QwenSparseAttnBackend.forward_decode", "original_physical_page_ids": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/728abf9804b05188f44c0b8973f76bbbce81a95c01149d641567bf0adced679e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "qwen_next_256k_long_output", "scheduler_mode": "decode", "scheduler_sequence_lengths": [208582], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3973, "uuid": "e899448f-1de2-5c76-8e72-63cb75d32a5a", "validation": {"axes": {"B": 1, "MP": 3260, "NP": 3260, "P": 64}, "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.07243998814374208, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e899448f-1de2-5c76-8e72-63cb75d32a5a"}}
 {"actual_requested_output_tokens": 7948, "adapter": "Retain all live pages byte-for-byte and remap their page IDs; omit unreachable allocator pages.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 204608, "label": "256k_long_output", "locator": "requests[74]", "model": "claude-fable-5", "output_tokens": 7948, "reusable_prefix_tokens": 204224, "reusable_prefix_tokens_256k": 204224, "role": "main", "selection": {"field": "output_tokens", "note": "Long decode trajectory within the Qwen model context limit.", "percentile": 99, "target": 7948}, "session_id": "1b577bfa02cd2c421c0b59c5cac702b3c287", "t": 3693.72, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 16, 1, 128], "axes": {"B": 1, "MP": 3322, "NP": 3322, "P": 16}, "cache_predecessor_locator": "requests[72]", "capture_decode_steps": [0, 1, 2, 3, 3973, 7946], "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "engine": "SGLang", "engine_block_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/54f2bdffa34c41890a61a6d8217ab3c652e9bb81ffde4e6b497b014d3f172ce1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/524150f85d7fc815b296bba4da027f8782a5347ff4a1702cac1503c22ac3913d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "93a0fe2aa6d6a9252367be8da47bf2577fed128d48c71bd4d3371f3c8f22aa25", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QSAIndexer.select_decode_tokens", "original_physical_page_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/2d9f6c8e79d56cc512b2d484a6eda8bde1b6b7c4f80a4bbb93645f60952d46e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "output_adapter": "Serialize the engine's selected set in increasing logical-ID order; preserve membership and -1 suffix padding.", "padded_query": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/9f87d7a06e613f7a542dce2c8653d115dd932c1d8d6e361393bfa97734537011.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "query_adapter": "Remove verified zero query heads inserted for the backend's eight-head tile.", "request_id": "qwen_next_256k_long_output", "scheduler_mode": "decode", "scheduler_sequence_lengths": [212555], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 7946, "uuid": "1a06c89e-7764-5a29-8664-4f1ee515e676", "validation": {"axes": {"B": 1, "MP": 3322, "NP": 3322, "P": 16}, "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "device": "cuda", "max_absolute_errors": [0], "seconds": 0.0043017249554395676, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a06c89e-7764-5a29-8664-4f1ee515e676"}}
 {"actual_requested_output_tokens": 7948, "adapter": "Snapshot after the current-token KV write; retain every live page and remap physical IDs.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 204608, "label": "256k_long_output", "locator": "requests[74]", "model": "claude-fable-5", "output_tokens": 7948, "reusable_prefix_tokens": 204224, "reusable_prefix_tokens_256k": 204224, "role": "main", "selection": {"field": "output_tokens", "note": "Long decode trajectory within the Qwen model context limit.", "percentile": 99, "target": 7948}, "session_id": "1b577bfa02cd2c421c0b59c5cac702b3c287", "t": 3693.72, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 64, 2, 256], "axes": {"B": 1, "MP": 3322, "NP": 3322, "P": 64}, "cache_predecessor_locator": "requests[72]", "capture_decode_steps": [0, 1, 2, 3, 3973, 7946], "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/5a28c0b8b8a79b1bcead6f9ea2644a7eed213f5110ffcaffc855a8185c3ad25a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "93a0fe2aa6d6a9252367be8da47bf2577fed128d48c71bd4d3371f3c8f22aa25", "layer_id": 3, "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QwenSparseAttnBackend.forward_decode", "original_physical_page_ids": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/2d9f6c8e79d56cc512b2d484a6eda8bde1b6b7c4f80a4bbb93645f60952d46e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "qwen_next_256k_long_output", "scheduler_mode": "decode", "scheduler_sequence_lengths": [212555], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 7946, "uuid": "8827697e-7478-5d3e-b017-ca652334a3b6", "validation": {"axes": {"B": 1, "MP": 3322, "NP": 3322, "P": 64}, "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.07027529133483768, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8827697e-7478-5d3e-b017-ca652334a3b6"}}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 128, "TopK": 128}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4", "validation": {"axes": {"M": 128, "Nkv": 128, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.014219233300536871, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4"}, "variant": "rows_128_compression_0_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 128}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/4b208d445b76a9e22b7ba8aeb15462cabf8a2fb8fea232fa6b891071e1bd855b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4", "validation": {"axes": {"M": 128, "Nkv": 128}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.014219233300536871, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4"}, "variant": "rows_128_compression_0_cached_False"}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/cab42d02ef8e1bdab85d77ae269c85693eb1812e200373e99b3924665d385f2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/c266a1c92d4f6cd3062553bb9c99ff9db32cdade2ad7c4be4424a6614286b082.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a9ef4b85ed35d2174174084a725cccf38c9ef6407f8b937b32ba67ba5cb57acf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7531a7c409a876f61ac259c161074206c934d45aedb50b4ea309d8ea1a80e55b.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7414061c-d565-5448-9e7a-78d5b1ebfdb4", "validation": {"axes": {"M": 128}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.3113021850585938e-06], "seconds": 4.782656603958458, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7414061c-d565-5448-9e7a-78d5b1ebfdb4"}, "variant": "prefill_128"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 768, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 768, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 3.690998049918562, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f"}, "variant": "prefill_128_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 768, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 768, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.7128978883847594, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503"}, "variant": "prefill_128_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 192, "TopK": 640}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70", "validation": {"axes": {"M": 128, "Nkv": 192, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.011357225012034178, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70"}, "variant": "rows_128_compression_2_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 256, "TopK": 640}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9", "validation": {"axes": {"M": 128, "Nkv": 256, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.010068903211504221, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9"}, "variant": "rows_128_compression_1_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 768}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/274614eec2f892ebe8291f84a02c8c52442cf7fc51a0bc0b7244c82b4193f1d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f", "validation": {"axes": {"M": 768}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 3.690998049918562, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f"}, "variant": "prefill_128_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 768}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6124ed8165865b9abde8ddb3856758f856b53c2b64afc98302b5a615d0e451e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503", "validation": {"axes": {"M": 768}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.7128978883847594, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503"}, "variant": "prefill_128_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 192}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/5951560539c31667434e9faf5c9f0c9e7f5e7cbc32296c43ab32eef581c10d3a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70", "validation": {"axes": {"M": 128, "Nkv": 192}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.011357225012034178, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70"}, "variant": "rows_128_compression_2_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 128, "Nkv": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/ff38c3cc6092e80e32b3e1ec3adb2fde8aab515b6241d22dcb7752f9fbe76a50.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [128], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9", "validation": {"axes": {"M": 128, "Nkv": 256}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.010068903211504221, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9"}, "variant": "rows_128_compression_1_cached_False"}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f18622b409563c96228e8293a5140a6bad27db5cd1ec4521702671c4948fbcbe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/875157d7a4b63732f005fccc03c4c154e9e1b4760317fe3c658e8fee3014dc4c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/101a87b53a0f82bc273d8c226c0d729c276ecd4faa4fde842f8cd968f676402c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/41f7eb2d843f705dfd20f53ad4d3220934c0020a5dd2852f221cf0d63daae43e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "712fdc6e-f456-507f-9fa3-dbc410c44cce", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 4.470348358154297e-08], "seconds": 0.00314719183370471, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "712fdc6e-f456-507f-9fa3-dbc410c44cce"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5481769633479416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b5374401-90e6-5268-96c8-b315c811aa64", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5974064632318914, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b5374401-90e6-5268-96c8-b315c811aa64"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3613db7e6c03bf5b77e75bf0f05a67f5bf71a1ef12e59c0d65e05a1c695a25ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5481769633479416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a9ff30c9fe5aad7f4f45297055b4c7df9736e9ed160969acbde0e74f589187d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b5374401-90e6-5268-96c8-b315c811aa64", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5974064632318914, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b5374401-90e6-5268-96c8-b315c811aa64"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/639c79b0fec07e42b02799a5e8c2136dbfab449686e3b5c21c734566f0c2f23f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [129], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "a141972f-4cf7-5111-967c-2dcb6463b86c", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.07541759125888348, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a141972f-4cf7-5111-967c-2dcb6463b86c"}}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/541ac6d7956b18859b40d4f8393df652ef15b6d596cd2de588fa1c9a291fbab9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a550e7a4da728acbf6496ca2f1dcd7e77204e926b6109ae241c6c7281db19d97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/54d2b115bf542bcf4433f9eb38481ee78b79522851f57ed8839d086ce54c1cf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f3db39033001d5aa6ded5a26ceca2bb4261d8f0b52ef8de213adfd7d1370e373.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "5d9a5db7-f338-578b-b8b0-8d3d1e5a06c8", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.002588692121207714, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5d9a5db7-f338-578b-b8b0-8d3d1e5a06c8"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5461130570620298, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5964415301568806, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ae27b9f4145d74b3bf24000f1cace3128d66139de78b6b512460d616d11012b0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5461130570620298, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72d7d3d27fbf96252492fd0d82886fbe79bcd66fdd5ca5e40aed69435c0e37cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5964415301568806, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/a8c15c2892df53a2ab99563f6897d7d423bdec63e23a36d65f2426bc50c872d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [130], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "eafd3828-d238-5d93-ad23-6443cd632c0a", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003597497008740902, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eafd3828-d238-5d93-ad23-6443cd632c0a"}}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/451565d54bd2a077dfefd6266d8eebf60da524a1aab2bd30316b03a2124c3c72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/06b06c7ef0de7911f410a31395090d18c9385704026e95428b51d3a33f55468b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/eff5114fd7e82fed274487ad2961094ad2b29e6e1605b17825fe01992842a1b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/930cd3c1337652be54b78d92e8b4529f05de93451f2c67451a1683b52e76d70a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "ea1d4f58-48fc-51af-8df3-e239f1fbddf8", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.001954874023795128, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ea1d4f58-48fc-51af-8df3-e239f1fbddf8"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5454176180064678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.4939532349817455, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b725efa8c9b438dfa626b9811fae2be2dd2cf2be8686c55b1d2c2a3eec425346.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5454176180064678, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/adb06105a0fc4e5640b89a12d54d3dff52967c9b9cd39d0b9d30b78c28cab534.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.4939532349817455, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/ee0c3e5977b92eb9861bf4f0823a3eaf89448822a51a43cec93c61f6962a5765.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [131], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "3aa1fefe-5390-5840-a649-b5e50c2ffd2f", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003529265057295561, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3aa1fefe-5390-5840-a649-b5e50c2ffd2f"}}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/626a5d5850cc927b7b73f10a1efca1cadf7438b266d24d4482cc15fc547c5c73.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/e0f79bac5b2e2fdcd3f62e3f6b9dbb5487c660c376405b786a15f5c7a6240533.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/fcae52528ea4033fdb8d8944bb39c26a337caf7f332e1d99b28c61ee0454fec3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/74a7a774ef2576f0fe5dc5efdc0a7ec213cc6a1712997493f5b4d412776dd47a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3e2153a3-11cb-5f0a-bedf-28f7e1b35d27", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0019683837890625, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e2153a3-11cb-5f0a-bedf-28f7e1b35d27"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5454356167465448, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "9502d66f-df7e-5068-96cc-e538986fec52", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.346023241057992, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9502d66f-df7e-5068-96cc-e538986fec52"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975cca1048447dcffb531bd0614cb335662954a0c8c406f7dc0c8e22146b8468.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5454356167465448, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/773e8694bc4562765e8352be7d14779de9d03c79fbff751c47bca0ecf25ad732.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "9502d66f-df7e-5068-96cc-e538986fec52", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.346023241057992, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9502d66f-df7e-5068-96cc-e538986fec52"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/1508164425014a88077326c657593b16b60054b519619b3dc450045b09fa270a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [132], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3d3493ba-7f57-54e3-94c5-34e37f52057a", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003510093782097101, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3d3493ba-7f57-54e3-94c5-34e37f52057a"}}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/e32a6f2a5dee7bf94613d1119eb93728c2fece3a45f1376a14e0fcae0619dd08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8961bb7b9822d25fd9740551eccef03d4de9c37e5a3fa77d2f478fb41c4705d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/fa13a91791f94fbb3576517c9ba62a7bc7a050ca246f54116da8cdd24a749719.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/aafe354428fe8f4b4a5786fa99a3314b16ac6da164c097d35fd959d91b4d6d2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 52, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "af60e984-b386-57aa-9a1a-447684a59bd1", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.4901161193847656e-08], "seconds": 0.0018545021302998066, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "af60e984-b386-57aa-9a1a-447684a59bd1"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "selection_step": 52, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.5513619473204017, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "selection_step": 52, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.34558157389983535, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6ec2fdbd40852627ad40f6e45bb1861784e13f7603e921fa695b407a6651e541.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "selection_step": 52, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.5513619473204017, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9b50d91c80b7193668ca425bc833a2ca67830e7940931f2eea811d7c06cfffd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "selection_step": 52, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.34558157389983535, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7f3d60ab4b9815d9a00abe6dd4d87bf56c50653ef4866737f18408e0a0a7d6e4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [181], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 52, "uuid": "04f8298b-720e-5f9d-be0e-4318bf74de22", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0034822807647287846, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "04f8298b-720e-5f9d-be0e-4318bf74de22"}}
 {"actual_requested_output_tokens": 107, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2fa8dfa782ccb71979b7e78b45b04993c59fea59271cf633354c733f27a64540.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/af59d6504a76e0e09c7994c45327914803ee7e1b15b417758e14ecce2491341c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6a332a1c9104697eb316448e98aaa24eb906ec902529dd4a53edcc9419e4260a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/e94d8ec9d55a8545c5891c9e09175b3da49409b494e9bacbccfbdaee082d67d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 105, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "99c44408-2bf6-5b21-ac9e-d9d49c5eab60", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018702130764722824, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "99c44408-2bf6-5b21-ac9e-d9d49c5eab60"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "selection_step": 105, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "900885d5-e628-566f-9f35-d67aa7a15924", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8859752179123461, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "900885d5-e628-566f-9f35-d67aa7a15924"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "selection_step": 105, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3394766910932958, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7ee77b2d5eaa8d59523f4c4fd6a90e761b356ca210a1cd3eb96cf1c8a8f6df3b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "selection_step": 105, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "900885d5-e628-566f-9f35-d67aa7a15924", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8859752179123461, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "900885d5-e628-566f-9f35-d67aa7a15924"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 107, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d7793b41566da8ff92174b4fe317f48bd126dd0285883275d88cc1d866274b4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "selection_step": 105, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3394766910932958, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 107, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 128, "label": "short_request", "locator": "requests[22]", "model": "claude-fable-5", "output_tokens": 107, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Shortest recorded nonempty-output request; sparse-history boundary.", "percentile": 0, "target": 128}, "session_id": "16c18d979235980e8021cfb0a0b3b386305a", "t": 1452.253, "uncached_tokens": 128, "uncached_tokens_256k": 128}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 52, 105], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/57d71f41692fe6a85ea2572ea2af77bf7a38d4e21377816da3182fc2cccdc327.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "5729f5afa5cdb1661ab1322da2559cdc1e4e03e927be5481169405ed4b4ce9b6", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_short_request", "scheduler_mode": "decode", "scheduler_sequence_lengths": [234], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 105, "uuid": "1ad1d4ca-c265-54b0-8bcf-123948e84b86", "validation": {"axes": {"B": 1, "N4": 1, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0035882792435586452, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1ad1d4ca-c265-54b0-8bcf-123948e84b86"}}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 2944, "TopK": 128}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b", "validation": {"axes": {"M": 2944, "Nkv": 2944, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1716800257563591, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b"}, "variant": "rows_2944_compression_0_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 2944}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/a0c2808625627f1d5867bfcc25ee7419f018ba66425d689f6e7b7405d53525ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b", "validation": {"axes": {"M": 2944, "Nkv": 2944}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.1716800257563591, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b"}, "variant": "rows_2944_compression_0_cached_False"}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5d8c3e4fdb71cec3d2ac6c09410eb5085007af7186be36f045e7e1a18e6d9ede.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/98ff14aef32cc15b3569c3fd0cd26f94ec87d4142f00423b731832b0445968a7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2b8a4cef93226682381aaf5a1f95dd30ff0a543068bbb6d025169e2ba3be244c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/c717e32a02b46c9a2dd4cf566674122cc6342fafcea2c9f7ad4e5d8103ed8b29.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "01578ef1-d67e-5840-b8ce-c8a26ccd7515", "validation": {"axes": {"M": 2944}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.2186508178710938e-06], "seconds": 0.008137943223118782, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "01578ef1-d67e-5840-b8ce-c8a26ccd7515"}, "variant": "prefill_2944"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 17664, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 17664, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2196662919595838, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1"}, "variant": "prefill_2944_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 17664, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "acae606c-4112-548b-b870-cec7314a3f82", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 17664, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.5008114967495203, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acae606c-4112-548b-b870-cec7314a3f82"}, "variant": "prefill_2944_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 4416, "TopK": 640}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a", "validation": {"axes": {"M": 2944, "Nkv": 4416, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.16766805900260806, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a"}, "variant": "rows_2944_compression_2_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 5888, "TopK": 640}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "52560299-af35-5351-a4f0-cdcf97141e87", "validation": {"axes": {"M": 2944, "Nkv": 5888, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.16742154397070408, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "52560299-af35-5351-a4f0-cdcf97141e87"}, "variant": "rows_2944_compression_1_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 17664}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b3cfbc03b72048cf44ed8a4039e786b73b76e95e4609b535301f8d9a5312ed5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1", "validation": {"axes": {"M": 17664}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2196662919595838, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1"}, "variant": "prefill_2944_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 17664}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c68b608397eb24c9b125dad1eb246e8061b89dd9dfdc001f5f3e21f4a0a5afc1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "acae606c-4112-548b-b870-cec7314a3f82", "validation": {"axes": {"M": 17664}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.5008114967495203, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "acae606c-4112-548b-b870-cec7314a3f82"}, "variant": "prefill_2944_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 4416}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/0bacc9640c92375c4b994168ebb73235f44df4b1beabb763ef988c8f9a6892aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a", "validation": {"axes": {"M": 2944, "Nkv": 4416}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.16766805900260806, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a"}, "variant": "rows_2944_compression_2_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2944, "Nkv": 5888}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/6417382a44a9c9038a2bac400f59f12b9a5d622134f9ef62ba2cd47e1d70d8df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [2944], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "52560299-af35-5351-a4f0-cdcf97141e87", "validation": {"axes": {"M": 2944, "Nkv": 5888}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.16742154397070408, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "52560299-af35-5351-a4f0-cdcf97141e87"}, "variant": "rows_2944_compression_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ecbb2abef863d2c678d2a73ea15c92c4da840f15f15a2f12bc555bb1a1b2f537.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/53274d3b8f9e5296f5391b03e24829abe888b0f5db1f2c03ffc91ba0586a8e42.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/3512952dd2bb124e7766f7bcfe8ff20f3a72a39ad2253b639e385a64a16f5ce6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5a8a3b812929bb630e5d81e36c94f640f353d61dbaba4c9df0aa27d3ee9fdc02.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7a7cdbd4-6374-51f1-9f99-53298395c41d", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018858392722904682, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a7cdbd4-6374-51f1-9f99-53298395c41d"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "88e96974-46dd-55bf-a772-be150b017d9a", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8851861101575196, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "88e96974-46dd-55bf-a772-be150b017d9a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33914247201755643, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2bfc386cf723d8edebed0e10705c122fb4b29573de51af6b14f934b41691b491.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "88e96974-46dd-55bf-a772-be150b017d9a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8851861101575196, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "88e96974-46dd-55bf-a772-be150b017d9a"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5cc11b258e00ce827774802b27b16de4798d35894d5266b4a48a76dbb5c8f687.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33914247201755643, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/3c4a0109de069cb70336437c4a360dbe30b205f46a5660ecbbb36f8721957fe9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2b87464912950073b063c001cfeda4bf4f6e105ba4d22fce2437368c298436e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d715c1df4e5f698fbf31d7c76c7a47b5b34d44e359932d58f91f7014b22c249e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2945], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "2c055caa-e9f7-575b-9d5a-36376e19011a", "validation": {"axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004787247162312269, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2c055caa-e9f7-575b-9d5a-36376e19011a"}}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8a09524e6a0885e674491813771a97ec342439532070cef8bce961c86064b318.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/41667d74739c9325aed400cacf15357da10c9abd514c3994fcba4ea068d1beb5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f67ef9cecd7a5614930dba5c720a9e6bbb60a1352992ae77d7395793f0e7e6b0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2db5bf2fae7e10ab42417743f036c4833075f1b405995419685d0a769645cb0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "2898ff25-4c56-50ba-ac84-59f2d0d11dce", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018871850334107876, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2898ff25-4c56-50ba-ac84-59f2d0d11dce"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.882862709928304, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3401966840028763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/67be217d1f1b345b7acb770cb15b68678df3ddbce7a96616fa2611fc66760663.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.882862709928304, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4bb3d95910e0d13a3c06c3150702fc50658ab733b73b335145eabc7b092d81.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3401966840028763, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7efc14437796e3a9def741b8ccfd23401c5d3667f5491736dbf67b3488f9a822.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2b87464912950073b063c001cfeda4bf4f6e105ba4d22fce2437368c298436e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d715c1df4e5f698fbf31d7c76c7a47b5b34d44e359932d58f91f7014b22c249e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2946], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b703462c-a066-552b-9a2a-a557dc484675", "validation": {"axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0037732706405222416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b703462c-a066-552b-9a2a-a557dc484675"}}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/d82afffac00ef0f037bcfe89fd81a5aa8501383643f4ef9d4563a8e806359b5d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6b24db472c6f5ca2d7d0c6a38f4b6b014ac4e5a0dc18498f998ac9b76df0bf6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5fcfafa715b296db1f98fc06e4ba1091e495d8d59bf1ad22fbb05076527c8900.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/1e22412c060d2e6de86dd06c132b7edbd50dd3fd3680347b61282959d8da45a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "dc6ed9ab-498d-55e0-a719-90f442212932", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 0.0], "seconds": 0.001883982215076685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "dc6ed9ab-498d-55e0-a719-90f442212932"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8487191488966346, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3383104531094432, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5351c07b703e638b20c7d6996f101df7a16b127fb7994fc75773898a467184ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8487191488966346, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eed89bf1258cf329af1120082ad65c909893c07839f19a99b5e84f0ed3b98f0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3383104531094432, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/6c3a5f9e70234d34d9119bf98227965d1b5c1eb018cea6e4cd698a0bbc6313fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2b87464912950073b063c001cfeda4bf4f6e105ba4d22fce2437368c298436e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d715c1df4e5f698fbf31d7c76c7a47b5b34d44e359932d58f91f7014b22c249e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2947], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "63d076d7-258b-56e0-acb9-9d2094dfa87e", "validation": {"axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0036922399885952473, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "63d076d7-258b-56e0-acb9-9d2094dfa87e"}}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2611a26297b6b8a7b0adb4d2cef7c1a092013fca09cac8882c076480c55f1c01.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/aa4f10ef3d2ee096f139ab939dcf966423dc00a4e439aa5a8ca255b4f7471381.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f65205944f2852ab00b0076eef054e616596420f862a5fb59787defea278f832.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a9fd9d2d975cc65d1eda502c562c3130c39c7e715aef8f452bf80204c361e820.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "b94d7d4e-92b8-531c-a7cb-4e312a5dd352", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 0.0], "seconds": 0.0018848809413611889, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b94d7d4e-92b8-531c-a7cb-4e312a5dd352"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5414832346141338, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3382181888446212, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c6c1d66c8d69b3e1a2cfde8ce79ae13348f892aec9663beebf5019ae700cbc5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5414832346141338, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b7f05bb25649cb16ab8b635bddaba5c0f36e289b1738d1bb34344121648d062b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3382181888446212, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/12407dc271b16d2ed0a6ba44d105332c0fd4d0c1bbe27e5538e9eb5d314a0bbd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2b87464912950073b063c001cfeda4bf4f6e105ba4d22fce2437368c298436e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d715c1df4e5f698fbf31d7c76c7a47b5b34d44e359932d58f91f7014b22c249e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [2948], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "5df6a182-3f20-5ae9-80fc-9dc2b9171d37", "validation": {"axes": {"B": 1, "N4": 12, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.0035967896692454815, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5df6a182-3f20-5ae9-80fc-9dc2b9171d37"}}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/634c5bb6d35b04bd9760f9bf9508fe40309bf57057a39fb8887eaf3ef99804a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9b257c83fee44304bba5cf11cfcff63a9979a901fe449763f13fde87ca417d5b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b1e442a7b0cf6eca71ed385537595f533e7827e4a327bb65abb6f972211042fe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/4a0fda3b914dd8479103eb1959bc87c69a6700780afa635495a3e39b34d920e9.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 905, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "3429bfab-7670-5910-9da7-b80768819de5", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018188119865953922, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3429bfab-7670-5910-9da7-b80768819de5"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "selection_step": 905, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5399576942436397, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "selection_step": 905, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "626d1307-1493-5593-8079-d1a58b9008e1", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3377021881751716, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "626d1307-1493-5593-8079-d1a58b9008e1"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cb66ac4f0cd1c5175891393aaf6535b8959dbb01454a0d8a74785c0088566473.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "selection_step": 905, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5399576942436397, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/084c707685328192664f58d19c64af058073664f9c61cf530648c508e9fde818.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "selection_step": 905, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "626d1307-1493-5593-8079-d1a58b9008e1", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3377021881751716, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "626d1307-1493-5593-8079-d1a58b9008e1"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 16, "N8": 2, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/a516db75e2c91026984ae37283d8ca6f11d91b7567c7d5ac5762d3bb26dc01fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d265caf29c1023c85155dffb8e74f26c777f56e3113ed62c77955e8411c0c16e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/73867bdd1fc6cf6a8e969490f87a8578cc0bf91702be620e48996d51b695bcf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [3850], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 905, "uuid": "67c1d03b-d10a-53eb-bf5e-34df0f9df57b", "validation": {"axes": {"B": 1, "N4": 16, "N8": 2, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003837081603705883, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "67c1d03b-d10a-53eb-bf5e-34df0f9df57b"}}
 {"actual_requested_output_tokens": 1813, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/09ccd29f0829f65025190c659895407160ecff9b3acd63ea211290aba52e59eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/67c842f238502711c6a5adc42ccec5aefad2c7a0c51351b64ad2b17989cf8c63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/21c20f4cd20758f90ac00e0362996652d00ef5c67ad1f3fb19c4a96a72f52c94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2f6c9cb96b41c1b7f56409108ecf4d37389b65450d99fdc3474edf33cce33010.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1811, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "36754b68-5dc4-5c2a-9dbf-c57ca9a6c523", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018437537364661694, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "36754b68-5dc4-5c2a-9dbf-c57ca9a6c523"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "selection_step": 1811, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5410367511212826, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "selection_step": 1811, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33674867870286107, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92401f14b7149adf55c04a902effbe92f95f3067fd98764403cdbb134c5e3c57.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "selection_step": 1811, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5410367511212826, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1813, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/36452df18abf87401ce997afb6874ccde11cbad2811c4c944a3846b3c67c0d63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "selection_step": 1811, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33674867870286107, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1813, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 2944, "label": "cold_main", "locator": "requests[137]", "model": "claude-opus-4-8", "output_tokens": 1813, "reusable_prefix_tokens": 0, "reusable_prefix_tokens_256k": 0, "role": "main", "selection": {"field": "input_tokens", "note": "Main-stream request with no completed-request prefix available.", "percentile": 50, "target": 2944}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 63769.496, "uncached_tokens": 2944, "uncached_tokens_256k": 2944}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 19, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": null, "capture_decode_steps": [0, 1, 2, 3, 905, 1811], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/4d6f98e34f23db26f66aba205dbf952f4f17e92ca65bcbeb146d8a9f0c7503be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "29f4cb73ab5e5b4efbca25f5c100876d73d535f9b7ef3856b82762f3758891e8", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/27b024bab8e0282f4d4a168257a72277b908f11d284e5eb312bc127b9e1f2c4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f07204d57b7e50c1a71e4c5f5b25bc5858c3dd21ef9fcfb12f9a820763e89677.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_cold_main", "scheduler_mode": "decode", "scheduler_sequence_lengths": [4756], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1811, "uuid": "ee56d2be-7be4-5091-a4e6-d827a6c6c4e3", "validation": {"axes": {"B": 1, "N4": 19, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.003902407828718424, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ee56d2be-7be4-5091-a4e6-d827a6c6c4e3"}}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 575, "TopK": 128}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b", "validation": {"axes": {"M": 448, "Nkv": 575, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.029641964938491583, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b"}, "variant": "rows_448_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 575}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/2e9515ccea2d768065a2df3fbf622d477617ce2c04d40ee5c323196fd25a0930.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b", "validation": {"axes": {"M": 448, "Nkv": 575}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.029641964938491583, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b"}, "variant": "rows_448_compression_0_cached_True"}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/31fc42268045ee1e94618e370c0ac5602e9cb8b05f42b0cfe6ddd6055b092b0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/679a3b46f81cbf1063e07f9259961c967556d9c4b8f7b410e59c65bf5f7ac659.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a9fa3b23fcbae3c06c3971f0f55d8237b43c0c98b378132eb5b90242f78d34a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ab3c19d5061aec6ec16d200ca91d4beda78f4937e22d6953857dcc5f59b3adba.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "3e29c0bc-7220-502f-b98f-d12f5720076c", "validation": {"axes": {"M": 448}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 9.5367431640625e-07], "seconds": 0.0030147982761263847, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e29c0bc-7220-502f-b98f-d12f5720076c"}, "variant": "prefill_448"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 2688, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 2688, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.71800921484828, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8"}, "variant": "prefill_448_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 2688, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "9972c26f-b34d-56a2-a043-a50684289467", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 2688, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.43521131109446287, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9972c26f-b34d-56a2-a043-a50684289467"}, "variant": "prefill_448_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 71583, "TopK": 640}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b", "validation": {"axes": {"M": 448, "Nkv": 71583, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04861454013735056, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b"}, "variant": "rows_448_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 142591, "TopK": 640}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1", "validation": {"axes": {"M": 448, "Nkv": 142591, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.08206645119935274, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1"}, "variant": "rows_448_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2688}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3fbf5e26e4521d9991708f9fb19f4195411b7cc38993ff921c4a485d05f81b5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8", "validation": {"axes": {"M": 2688}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.71800921484828, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8"}, "variant": "prefill_448_cached_True"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2688}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/721e2d71ad17d9562ecb9b29fde53181a1269552fffe6c1a661f0e0ce1409883.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "9972c26f-b34d-56a2-a043-a50684289467", "validation": {"axes": {"M": 2688}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.43521131109446287, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9972c26f-b34d-56a2-a043-a50684289467"}, "variant": "prefill_448_cached_True"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 71583}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/5aa0f0f71df781da73ddfac2f89060dafd5e25c31704012443ab3ab4b32c628d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b", "validation": {"axes": {"M": 448, "Nkv": 71583}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.04861454013735056, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b"}, "variant": "rows_448_compression_2_cached_True"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 448, "Nkv": 142591}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/16be1cf36a53699d6833d31f11c3c0fed1949831db510d9213d77c0138126018.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [142016], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1", "validation": {"axes": {"M": 448, "Nkv": 142591}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.08206645119935274, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1"}, "variant": "rows_448_compression_1_cached_True"}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/09ccd29f0829f65025190c659895407160ecff9b3acd63ea211290aba52e59eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/531586492d553c2e5470fea35d1a8e71c1b51640c9eb7cf4d382bdee15b35fe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/21c20f4cd20758f90ac00e0362996652d00ef5c67ad1f3fb19c4a96a72f52c94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/124078be107aa872230628cf9f1a4bf712639dde8284a56d8e83a208cf77cd93.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "1465f204-f3a6-5d8b-9d52-b0fa13447170", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018743081018328667, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1465f204-f3a6-5d8b-9d52-b0fa13447170"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5414900947362185, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33660757867619395, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fad04446054828a3cef9ece6aff7d0241474b0a4eb63689b4cff5038f331f5a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5414900947362185, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/479a2dba0999aae37cce8d64d834489a278533abaaf9fa7fe7aa9433a2c080cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.33660757867619395, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 277, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/762550f7a0d62818904f9e9d03bf7501461ed421c5e8dc946d3251818721882a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/477c74a8854f9cb323e4c40c2c3e25b13832aec507be35c33d812da2288737e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/627ae435e59b7294f04d8545f05db67cfa01059e3e470743edf6b1156774495b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142017], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "ce862982-589a-547f-8b47-e68851a8566b", "validation": {"axes": {"B": 1, "N4": 277, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006942179054021835, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ce862982-589a-547f-8b47-e68851a8566b"}}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/3f35a46e1b71132aeaa26b11d9f8669870077ff92d60304db890f1071634252a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/703fc47cd09c1854850618732b90bcca5a8d5d830a4f6c9c6e1e03aa28c27a10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a1402d76991ae1204549d6a7eb6d6d246b2f775991166f6daa50849d38f7c38a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/15608b0c826f90f4ac9d43ca49f675270843db6b910ce9daf5a23c7a87bda982.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "32358a16-c098-54c5-b08f-2163b985863a", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 4.470348358154297e-08], "seconds": 0.0020304671488702297, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "32358a16-c098-54c5-b08f-2163b985863a"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "83936513-cfbd-5365-b669-4af3541aa223", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5406065662391484, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "83936513-cfbd-5365-b669-4af3541aa223"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3359463354572654, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/49cc025ea813cde0c10ad4a4975379ef399c942158fdd176c0e7939f262e108c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "83936513-cfbd-5365-b669-4af3541aa223", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5406065662391484, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "83936513-cfbd-5365-b669-4af3541aa223"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5407919f93a422116875fa0a8c24ab2b93e1aeed07ed94fb961f932e2aaa1676.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3359463354572654, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 331, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7ae5a3c5b106a4e7a9b372f1d2317d8da960ca152c84805528ea2e68ca81bb4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d741354eb5ed44132c55bb53a3fd21077042d16311b653c01ec2325ce9ea52f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/627ae435e59b7294f04d8545f05db67cfa01059e3e470743edf6b1156774495b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142018], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "f907c5dd-eac8-5661-a36a-56aa7a7aa6ac", "validation": {"axes": {"B": 1, "N4": 331, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007241962011903524, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f907c5dd-eac8-5661-a36a-56aa7a7aa6ac"}}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b7100fd06873521c15063e004041dbbb5ab115d41e088e104cee155978691689.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/bc0019a176d85791d2541854684221862122180d104b3062b1ba0d2eb7542176.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a6132df243d5e1e5089ec529cfaf84b6118a142894660255316af37cf9567a8a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6a61e6220c02044b23e773d7331ee02655c05e23abb9a7055e020a1eb20eca5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "a89c8fd6-0750-5bee-bdd9-4a04d254cba9", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.4901161193847656e-08], "seconds": 0.0018712240271270275, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a89c8fd6-0750-5bee-bdd9-4a04d254cba9"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5402844678610563, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3366624996997416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2f4dc6e6f2fb9e2485019da36bd339c44e6f16cad6a483f2e235ae4672540811.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5402844678610563, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/39643d5869435e494582a3713560e2c3cc632af93b21fc616085f463d2177ee9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3366624996997416, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 345, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/db6d05c680d878ae66bd14d7eca06b58227b6087a5be9e2b876eb5bff57d0ffd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/8c5b2304cf8c552e21d25210358fd1916d95ba6510c5d3f338f67bd178d240a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/627ae435e59b7294f04d8545f05db67cfa01059e3e470743edf6b1156774495b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142019], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "2aecb81c-d898-53d5-8556-3d0991d65651", "validation": {"axes": {"B": 1, "N4": 345, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007075177039951086, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2aecb81c-d898-53d5-8556-3d0991d65651"}}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/092125054f5770a954bd7ad7d3fd694693f4b7c5c25fb339ffecd4223772a89f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/73066f038df87216b0dd08d84218757d2aa84d4beaeb51ca4eeb075ed43f667a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/58d632efb31ec216cd82bb4c9ebeb11619463ef7851d9a1874049f8f36000352.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/655e0d275def327f90477d4a263637222b207fbbb5d5a066ff44f7bd737e9ec1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "26024620-0ef7-5520-8f45-41cadd789175", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 7.450580596923828e-08], "seconds": 0.0018882136791944504, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "26024620-0ef7-5520-8f45-41cadd789175"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5398206459358335, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.40999557124450803, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/00dc8e682857a357b5cb113b5bd2ecf711e4e009deaabefac7ad46f1f1cc32b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5398206459358335, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/095daf71d11aced296aa26e3263a5d636dd496efa299718d42355151c244ff9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.40999557124450803, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 350, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/aa34027ce19d1b5fd9b111ab9815d52e924063790b3bdd77de783f7fe6483589.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/982661c49712bfb972d47d3745f7bfaa9a70f3fd78ccec21fd483317224a57df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/627ae435e59b7294f04d8545f05db67cfa01059e3e470743edf6b1156774495b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142020], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "d2090849-b97e-5099-ab49-e5b472e4cc56", "validation": {"axes": {"B": 1, "N4": 350, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0067698098719120026, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d2090849-b97e-5099-ab49-e5b472e4cc56"}}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6d429a556db57aa1339ef11a90e4351d93168c625b0a70826e506c10d5f22319.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/10fc931b26250f3875a489a56de3b4231954a70ca5d0e100926b8eee25f495b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/23ea213014268f865ee574bdb75aecc5946545d58c3fadfc3e53c259724b2a8a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ec764eca7772d53199ec0489dadfdf5296b90f02f4def3ac55c97a0e9d04455f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 44, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "693a1e16-0aa3-50ad-a4a7-f7ad88f2ecd4", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.4901161193847656e-08], "seconds": 0.0018469691276550293, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "693a1e16-0aa3-50ad-a4a7-f7ad88f2ecd4"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "selection_step": 44, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5383333088830113, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "selection_step": 44, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.4002616498619318, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65206f1fb0f4e5b68da4b8550ad3945901c87c4c9a45f47a34a153182a7d9784.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "selection_step": 44, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5383333088830113, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/94d6a9ec2fe888116e1b24d9fca75f6d2acb842bbc6cc6e149a9ec1d6bc08649.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "selection_step": 44, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.4002616498619318, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 402, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/13259280168ddf8e913bf47e47dba00417d263046a052a1df51a20dbf9cdc1b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/b89a7d32aa5dc84f736377d8d13c816a56f35a7515fd9df8df73bcb4450730ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/627ae435e59b7294f04d8545f05db67cfa01059e3e470743edf6b1156774495b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142061], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 44, "uuid": "484ddd41-b31e-55d1-9f15-9ff33653f4ed", "validation": {"axes": {"B": 1, "N4": 402, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0072348639369010925, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "484ddd41-b31e-55d1-9f15-9ff33653f4ed"}}
 {"actual_requested_output_tokens": 90, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/11da69eaf8fefaab4fa858f5f3d2e3a31a27c3d2570c7a589bfc7b01786a085e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b9fd9bcef3855ece2b21f284a09335687b7c3015189164d11c518ccf3eb2a117.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ef4bf00a5b167d6644e3247d4c6a8ba0dbd2c1649ea4005a19da42480aba063e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2d2875093a95a79836259fbaf778c7d70884a4371ec8836305d746615f4ddca7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 88, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "68e00c17-4a02-5fe0-ac99-f0423af525f7", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.1920928955078125e-07], "seconds": 0.0018689977005124092, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "68e00c17-4a02-5fe0-ac99-f0423af525f7"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "selection_step": 88, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.576311593875289, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "selection_step": 88, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5318765640258789, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7dcd475cbfc559eec256077d51eb7a564e55b69562ea804805d36514186b5f5d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "selection_step": 88, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0009765625], "seconds": 0.576311593875289, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 90, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad8f02c0f6a4c101c7917b1fd4638e47bc2561d97a350d3bfb19a036ccec35b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "selection_step": 88, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5318765640258789, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 90, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 142016, "label": "full_context_p50", "locator": "requests[452]", "model": "claude-opus-4-6", "output_tokens": 90, "reusable_prefix_tokens": 141632, "reusable_prefix_tokens_256k": 141632, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 50, "target": 142016}, "session_id": "0470d446a4514dfe0c6ad0be92853bd13287", "t": 113074.811, "uncached_tokens": 384, "uncached_tokens_256k": 384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 180, "N8": 2, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[451]", "capture_decode_steps": [0, 1, 2, 3, 44, 88], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/6332feb5702e91e78917c8f10c44af16d94a0a07cbda7a546fde95e1e3097023.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "e262c52fc5a556a94e421a9cbbe377b90692e259758d39a27496d2929356cf84", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/b6dcd3df671f503fa224df6c2b65f7b5b9b10bc94c69f9f718eb95b5dbf9b0ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2cb6d4b7131c3d3129a07423fbcb290f03b47c72ba96b2dc271b844b95bcfa85.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [142105], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 88, "uuid": "680bd5dd-94fe-5dcb-84d0-9488c1db10d7", "validation": {"axes": {"B": 1, "N4": 180, "N8": 2, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.005677140783518553, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "680bd5dd-94fe-5dcb-84d0-9488c1db10d7"}}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d1c89357-aa68-53ac-b13f-62d900007022", "validation": {"axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.23019856214523315, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1c89357-aa68-53ac-b13f-62d900007022"}, "variant": "rows_4096_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.1952670449391007, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749"}, "variant": "prefill_4096_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.7681796588003635, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab"}, "variant": "prefill_4096_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 277247, "TopK": 640}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417", "validation": {"axes": {"M": 4096, "Nkv": 277247, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.30069290520623326, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417"}, "variant": "rows_4096_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 550271, "TopK": 640}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa", "validation": {"axes": {"M": 4096, "Nkv": 550271, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.8619707771576941, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa"}, "variant": "rows_4096_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 3583, "TopK": 128}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be", "validation": {"axes": {"M": 3456, "Nkv": 3583, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.19409079710021615, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be"}, "variant": "rows_3456_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4223}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/057baf4ac56eada3f421e3b654715f92f567b4ac36ec23076ed55c2c288b9dcd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d1c89357-aa68-53ac-b13f-62d900007022", "validation": {"axes": {"M": 4096, "Nkv": 4223}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.23019856214523315, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d1c89357-aa68-53ac-b13f-62d900007022"}, "variant": "rows_4096_compression_0_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b7ea650661faa21128e99ae8a360d1a6c8f3b6c2db71898712ba2bc6bada3fa5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.1952670449391007, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749"}, "variant": "prefill_4096_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/97d3501a043b033e123357b0360f2d08950c0e33dd6ee6e4cb19878a30be73fc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.7681796588003635, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab"}, "variant": "prefill_4096_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 277247}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8998d81445f99955cea4890a5fd74e14619fd0770a5e0e19c4bb4e66a308bff0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417", "validation": {"axes": {"M": 4096, "Nkv": 277247}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.30069290520623326, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417"}, "variant": "rows_4096_compression_2_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 550271}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/ea8c1db1cde54e4b71dd9162c3907e0d4f224226f7a1a598d87caeacb4a10e88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [546048], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa", "validation": {"axes": {"M": 4096, "Nkv": 550271}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.8619707771576941, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa"}, "variant": "rows_4096_compression_1_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 3583}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/9d6322037b8c795d5ee50cf7e3f27b2dd2d0e3093774361004c6bb0df96720ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be", "validation": {"axes": {"M": 3456, "Nkv": 3583}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.19409079710021615, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be"}, "variant": "rows_3456_compression_0_cached_True"}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7543e9bbce414d64a0732dba69ce43f0ea7ab43f1139f01d2dd1916385e922f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/57f3f5e2b28bac04085691a0a21c713f56790b54b280893e59f2ec76f752ec6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6ebf6770d66c58ac6fa2363663f29caa8d9f1e8be8b821863c9d2e50c415d564.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/1e889f9db33dc5b24ea22256ff74096ba0b82de45518c0ceb89eb7f3bfe1d08f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "de8359bd-2cfd-5d2e-8c9e-d78390c9b27b", "validation": {"axes": {"M": 3456}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.337860107421875e-06], "seconds": 0.008849332109093666, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "de8359bd-2cfd-5d2e-8c9e-d78390c9b27b"}, "variant": "prefill_3456"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 20736, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 20736, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.1965465568937361, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136"}, "variant": "prefill_3456_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 20736, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 20736, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.7529570069164038, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30"}, "variant": "prefill_3456_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 278335, "TopK": 640}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed", "validation": {"axes": {"M": 3456, "Nkv": 278335, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.24783538607880473, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed"}, "variant": "rows_3456_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 553087, "TopK": 640}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2", "validation": {"axes": {"M": 3456, "Nkv": 553087, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.3047167588956654, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2"}, "variant": "rows_3456_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 20736}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/996fe8e28b8a86d9c1f1867cc462932eab9d76bc83cc922a2685d83fb0cc454e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136", "validation": {"axes": {"M": 20736}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.1965465568937361, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136"}, "variant": "prefill_3456_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 20736}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c82585ec4ad6e3213842b4f8763d5a2541b688e8fc947613a5d8f18da3e92c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30", "validation": {"axes": {"M": 20736}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.7529570069164038, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30"}, "variant": "prefill_3456_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 278335}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d3cc24110494e5527024a63fb71ef9e595d30064dea53c65265cd31f0c7e3e8e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed", "validation": {"axes": {"M": 3456, "Nkv": 278335}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.24783538607880473, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed"}, "variant": "rows_3456_compression_2_cached_True"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3456, "Nkv": 553087}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/08b6a44f258777ffdf4677bfc0016845b942a995ddb9e4b983905ba457076025.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [549504], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2", "validation": {"axes": {"M": 3456, "Nkv": 553087}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.3047167588956654, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2"}, "variant": "rows_3456_compression_1_cached_True"}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b88a18b26979aa1a144dbb5bf1c92e630b6afd9c04cea9a76059d9daebc00bb8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/671453c01e9119fcc29636eb2d54b5ec6d11dc3c168db63006434db5b07289ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6fe606518897335d43fe744ea0eae34ef1ba71dbd26e3cdc5d182c35d9044030.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b945544d94def7d390c0be45633ffe15c4d5820515437ba7cab8ec98a5aa4c15.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "2032b687-a0fb-5150-a045-28fdfaea612a", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018634181469678879, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2032b687-a0fb-5150-a045-28fdfaea612a"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.876852915622294, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5329804020002484, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e9f5881ef740f4869a7d45e56ba53e3d6a926afee52ccd853b60e95fe4ec74e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.876852915622294, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dc8b8335236d1a531f0552dc5745a8b17a3647f65a5e66a1110fcbc9f474f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5329804020002484, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 369, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/0397b419e10c3671e7205c93fe56ce8eb081370ed3e253fa11240ec73c111b12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/8c1fb852b1d984db8a857b374e277719ee129fd98ef9db5b4dc88295dbdc91f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/43fa684a1d8e0bfc46e8fbc3e82a85a2c0ded168db4c0554bc8f95b007014cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549505], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b3caf995-3e21-582e-aeb0-c4ccf7c54692", "validation": {"axes": {"B": 1, "N4": 369, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006881288718432188, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b3caf995-3e21-582e-aeb0-c4ccf7c54692"}}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/452bed498a4492979528d657735f11b0aa8606a9d9633acc1c537ef9327508c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/49c3abb766f90072bc607266edde3fcb24b663a52a458d25bfe0b2dfc8476ba0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/872c33df706bd158a852d89c74321f841e5f3b64a4ccdc80212418efebb77172.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/17b4ca29db80556985ae2d2e435d3e64a39a4341d4a26d647d3e34ba6ed7f572.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "de10ba0c-dc65-5e21-9b85-b17198b0909b", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018859980627894402, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "de10ba0c-dc65-5e21-9b85-b17198b0909b"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8794198879040778, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5345803159289062, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/476ec669cddee30aa5b44b83b6b41f1c120c9debe41588de1d61cf9f7d5cb1cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8794198879040778, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a1ece55f8d06f4ffcd58b7314e09804e82dd07e73614a5cb0b80aa38f3901f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5345803159289062, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 388, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d1ce3d8c33a274908aa69eb9d482166f6ee1a3593d1cbb93f708674ce3b1fcb3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f786555415eb5fb2e376f57693883af7602b24814da9bd3f81424e9424e9eb27.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/43fa684a1d8e0bfc46e8fbc3e82a85a2c0ded168db4c0554bc8f95b007014cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549506], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "6184cc39-62e9-5827-9a9d-b8bf1f01226d", "validation": {"axes": {"B": 1, "N4": 388, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007385804317891598, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6184cc39-62e9-5827-9a9d-b8bf1f01226d"}}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9fe6c90a35151e65ebdc069a86b6064d7775fd9f494c150ec3dbacb9560f5a0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/75d7e1bc4e1302295cd0841822a9803a1d2765919f5f57eb66a80edaa2100c57.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/03505d5680fee2ee2e158f102ed1bd9f42226dafd4e8c11e4e7611c537a2d0a4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b241e4e05bf52529a13a111a188bd3b826fc6af20cc5a5bce3a80c22d6d71d68.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "ee51716a-39f7-5dae-9967-73e1d73ed8a5", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018774350173771381, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ee51716a-39f7-5dae-9967-73e1d73ed8a5"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.6839453438296914, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "8813329e-c415-5901-baa7-b9f423f06144", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5335805821232498, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8813329e-c415-5901-baa7-b9f423f06144"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89ec0981db984bc9f79f36faae81c7482efd69cabd51b856ae682d171b6bcf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.6839453438296914, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/eeee74431f12d103f1ae94bcb6a0c60863d6f0d67c75a7e98c1e8dc63283aa1f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "8813329e-c415-5901-baa7-b9f423f06144", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5335805821232498, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8813329e-c415-5901-baa7-b9f423f06144"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 448, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/5ea07399fb2809704213b49e477a34006908b308ab72e151baf8b43daa2adb05.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d6c58b288c10f6e25577024887246c2657e7dc5b8a3790059d216d4b77199308.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/43fa684a1d8e0bfc46e8fbc3e82a85a2c0ded168db4c0554bc8f95b007014cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549507], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "1cdf60bc-edf1-52ef-a731-dc5484a307f5", "validation": {"axes": {"B": 1, "N4": 448, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.007474483922123909, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1cdf60bc-edf1-52ef-a731-dc5484a307f5"}}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/587d50a1181bfc2e115d0d7679fe35343677b438ae18b8445097848be92fa88d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/86625bc365f40869284858790d370881c24d2ed78814af42969288cc92276cee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f9d27fe6521755633461dfe99c226446ec10aed05cd3f7df59cf53a6ea82347b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/3aa987c47d80e5080684dd4ab09baed9c3b807fc761551acdd3c6f29e0703a2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "7a8a85fc-f158-5260-9b7c-f03d75210337", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0019132322631776333, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7a8a85fc-f158-5260-9b7c-f03d75210337"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5393422758206725, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5224500061012805, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0f2db5c6d63f2099c0c0f335dd5861cfaf99bf7d27e8e4da3b4ac2addc7a1e4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5393422758206725, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/09bd9c9a9a58903fecbfdd12c9fb123e1f120750489f2716fb6d8ecd4a800438.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5224500061012805, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 441, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/af10addf0e19c685a1c78ca8a96e5b7ae33d5300b620c7fd1ce8f192bfc69fc0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/8d6375a2306a4ab79ecaa5443ca71b2a36297afcc41c7f9a544cc66164c1fa6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/43fa684a1d8e0bfc46e8fbc3e82a85a2c0ded168db4c0554bc8f95b007014cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [549508], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "c89ae7dd-6f4f-5366-8098-a7fd429fb63d", "validation": {"axes": {"B": 1, "N4": 441, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0076981522142887115, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c89ae7dd-6f4f-5366-8098-a7fd429fb63d"}}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/44d52cc60514bf928fd8d8d8ad8da816103b2bc754ba51261802d00d7087ac2d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b69c8ddb2d1363ad3185a3fc093203bbb8eb4346d5bf1ed0a1c42ff3fc65648f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/d9cf2a59a1f40e05bc61cfb389c41a977ee7294e2021d7e684ea4211bd2bf281.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b44fe0ccc9eb232715ed1e7658cd8ffdeb062f1b39535abe1d86dcb2c264a59e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3381, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "d52e2810-bf81-5ec5-a053-98a5d182a485", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018911552615463734, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d52e2810-bf81-5ec5-a053-98a5d182a485"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "selection_step": 3381, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5385668869130313, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "selection_step": 3381, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5277281040325761, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/46c10ee165b7b14dba209e5b6c246a00ca7393eecdf301049e4b1d813d750e15.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "selection_step": 3381, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5385668869130313, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/58ca62f583add82b057c3939488006f72162b2e1e75c899f2d6f6440bc66fbb4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "selection_step": 3381, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5277281040325761, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 372, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/b7446a392a91c18e6e2807e6822d4a7a450df70d531d273e62aafdc977c3fa07.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c259b7e48e226390555b8c79c52b5ab64b32fa536b090f7d31a7bfd2c292e068.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/dd4f3fe39c0fa38a5de8c79eb50d26084b6fa371b0496caa5831042dd3084a21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [552886], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3381, "uuid": "f1361c58-1268-51bc-9952-5b6fc88b9419", "validation": {"axes": {"B": 1, "N4": 372, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006781119853258133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f1361c58-1268-51bc-9952-5b6fc88b9419"}}
 {"actual_requested_output_tokens": 6765, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/400fe4c46a02e4b156dea951bc97354ebcf6ebaa35f6fb0a38290166e0aecb06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2d0a0c1f6559d7525675f4f2babaf4f71019dca88e376ef8796aa4b928752e8d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/093844abca9c400c9b1d582732b3b0e6dd6f1d178861e3275a56a54bac5628d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7b5d7cc57c7ba685d037f8ddd0324f90c10ea925cf8a2a65776c272d6ea6a180.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 6763, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "d2df577f-b8b5-5fb3-931c-54713f9be07d", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0019130720756947994, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d2df577f-b8b5-5fb3-931c-54713f9be07d"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "selection_step": 6763, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5377042600885034, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "selection_step": 6763, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "671e5f78-5429-5d4f-94af-24627c22937a", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3114455109462142, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "671e5f78-5429-5d4f-94af-24627c22937a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/53a57dd38899f4d36fca9da13688a2cfce6186b5bbeb1af2718ce5576dc81827.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "selection_step": 6763, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5377042600885034, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 6765, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/dbce240f22d5dac2b0d5db6e544f273c7b80f2f3a3cfce44c28ec12917b8d3b4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "selection_step": 6763, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "671e5f78-5429-5d4f-94af-24627c22937a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3114455109462142, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "671e5f78-5429-5d4f-94af-24627c22937a"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 6765, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 549504, "label": "full_context_p90", "locator": "requests[202]", "model": "claude-opus-4-8", "output_tokens": 6765, "reusable_prefix_tokens": 542144, "reusable_prefix_tokens_256k": 252608, "role": "main", "selection": {"field": "input_tokens", "note": "Observed order statistic at round((N-1)*p/100); exact recorded request.", "percentile": 90, "target": 549504}, "session_id": "21cde366f5bd2f0ff2031b0538496c7e2a05", "t": 10289.451, "uncached_tokens": 7360, "uncached_tokens_256k": 296896}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 457, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[201]", "capture_decode_steps": [0, 1, 2, 3, 3381, 6763], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/68e405c45f2767e885f3d77898cbbce68a4a2ccb6e99bd2897ed5d192afb177f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "58fb3d2530a18e97702acd19d1eb3cdf8c663f49a4024e7902d38a3fcef700d5", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/821a761023b4ad34649c0fb2421f5b5d8517b1ac1d99e5bb89c15855bbda2d65.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/43fa684a1d8e0bfc46e8fbc3e82a85a2c0ded168db4c0554bc8f95b007014cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_p90", "scheduler_mode": "decode", "scheduler_sequence_lengths": [556268], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 6763, "uuid": "aee264c6-9720-5a03-ba9e-f37fa8b448b1", "validation": {"axes": {"B": 1, "N4": 457, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.009693556930869818, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aee264c6-9720-5a03-ba9e-f37fa8b448b1"}}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 2559, "TopK": 128}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "8a834433-bbcc-5122-b239-be8569f77d02", "validation": {"axes": {"M": 2432, "Nkv": 2559, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.18988473201170564, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a834433-bbcc-5122-b239-be8569f77d02"}, "variant": "rows_2432_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 2559}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f4d5cf25fe219e3b4fff39f3cdda881e4a0d49e62f9fae3c28df320be2407871.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "8a834433-bbcc-5122-b239-be8569f77d02", "validation": {"axes": {"M": 2432, "Nkv": 2559}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.18988473201170564, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8a834433-bbcc-5122-b239-be8569f77d02"}, "variant": "rows_2432_compression_0_cached_True"}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/e06944592304ebe2e9f1e72f5cce6c6dfdd21e7216fe6c3e2dee72ae3a0647da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ae004ce47260c26518607c66ada50e743a52b0bd9158d58b3b214509bc3e8953.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/865bda57626b56ec75f993c9e1ba7702478bd83fe641d93c08283c15eb443504.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/d3beaf1af137bc48ddf030ed9cd9f39b8575880b1a16751894f667ae861fb442.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "6fd94728-92b0-551b-b4da-44400748dd68", "validation": {"axes": {"M": 2432}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.2186508178710938e-06], "seconds": 0.34629026986658573, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6fd94728-92b0-551b-b4da-44400748dd68"}, "variant": "prefill_2432"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 14592, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 14592, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.7905320059508085, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96"}, "variant": "prefill_2432_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 14592, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 14592, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.47346740402281284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e"}, "variant": "prefill_2432_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 497471, "TopK": 640}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a", "validation": {"axes": {"M": 2432, "Nkv": 497471, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.2923034420236945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a"}, "variant": "rows_2432_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 992383, "TopK": 640}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5", "validation": {"axes": {"M": 2432, "Nkv": 992383, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.3822329668328166, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5"}, "variant": "rows_2432_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 14592}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9209deabd296cef7f020391c422bbc7a19b17d47edbe91a97d18832b2e41ca4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96", "validation": {"axes": {"M": 14592}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.7905320059508085, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96"}, "variant": "prefill_2432_cached_True"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 14592}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e1d1b3fd9b31a86fa8489c8a90911af3124c1173ce8b375b4a3ee1ee4948e719.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e", "validation": {"axes": {"M": 14592}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.47346740402281284, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e"}, "variant": "prefill_2432_cached_True"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 497471}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d9dcb611421dc20fa83e7f9a3f6d6d14a85b90205c84e5150ce1aadbf559cfb0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a", "validation": {"axes": {"M": 2432, "Nkv": 497471}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.2923034420236945, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a"}, "variant": "rows_2432_compression_2_cached_True"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2432, "Nkv": 992383}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/2e52d1b85d76c5d98ff3f8de5ba9ed7928e5e0f952c153468778595e2fb3683f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [989824], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5", "validation": {"axes": {"M": 2432, "Nkv": 992383}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.3822329668328166, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5"}, "variant": "rows_2432_compression_1_cached_True"}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/044c34a6c301f1014ec89450c7ea729bf312b11f7a2014b354cda843e527d8be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2a03b86b6414c1205aea55f30c943e8988a75c9ace40e445ac0ead3fe011dd27.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6abc1a56e92245fd090e21c2a17e7e5dda6516a98108d9829270e8f6a1825ff1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/be80332ab064458e5a240a01c26b790413ad6eb90dad0b443c29b1157112ba49.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "288e8f54-fe3b-5db8-8133-df67f745a868", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.4901161193847656e-08], "seconds": 0.0018673227168619633, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "288e8f54-fe3b-5db8-8133-df67f745a868"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "fdc0d189-a357-57ee-b045-c08626f15054", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5379030527547002, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdc0d189-a357-57ee-b045-c08626f15054"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30503664491698146, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cf449316695599c75b7b2ee69cbb133c9dbb9f022175478bf30b6a6b6fbede4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "fdc0d189-a357-57ee-b045-c08626f15054", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5379030527547002, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fdc0d189-a357-57ee-b045-c08626f15054"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/323e1b7cc8b68ed487d74a5133f9220765ed192716b6ab11b7327abf177ae9ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30503664491698146, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 391, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d5515e0ae7b6e66ca2faa1758aed079363d31e087aabe24b345588f5996a665f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7a55658b2d19b9e8311dc79a5e3d51fa1296c6a17ce250d083ecea1f92a5bc3d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c328bfa734b290f1b4eb9308df7b36e60d09221fcfd313d5b8dff8254bc121a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989825], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "4f952a2a-90c1-58c6-8689-895422b92d70", "validation": {"axes": {"B": 1, "N4": 391, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.007083912845700979, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4f952a2a-90c1-58c6-8689-895422b92d70"}}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/14f79d894fa215c9339a45e31034cf053b0394262718ea5b2ef6f73b54ccbce8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/23b375470dc83c854e854da29dbd37b210429f265dabb2617a8d8dfd749df6d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ec30ba59e573958d9b848368683f7db1de30e14fa286d535215816e699cab27a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9064b5aa0898f9e54372f27d8f13a9da5713240e2a287800ae29bd33234a27af.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "7e4e5d7e-10bb-58f3-ab0a-0053dc4a910d", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018532858230173588, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7e4e5d7e-10bb-58f3-ab0a-0053dc4a910d"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5375021840445697, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "926dc757-1514-5c87-9773-99d219d9f473", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30510560562834144, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "926dc757-1514-5c87-9773-99d219d9f473"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c66df53e4e2aa9a78dc3478c1670b93772c7fcd09e9bc86a5b8dff72dfbb4b67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5375021840445697, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/267eaebdb6dc0081b729b82939fd95b36309a3731ed529e30205aca5f3117c28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "926dc757-1514-5c87-9773-99d219d9f473", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30510560562834144, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "926dc757-1514-5c87-9773-99d219d9f473"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 440, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/ff5eccd42de498fb60561c7370b15fdca4468368b975b6d7ec87ef54cca114a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/0f4aa9e73bbb46845c3094cef2c38793f32505cb1e26f4a01b8c33863e82ba6f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c328bfa734b290f1b4eb9308df7b36e60d09221fcfd313d5b8dff8254bc121a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989826], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "35baf826-1c46-56db-a032-0e97cc76924a", "validation": {"axes": {"B": 1, "N4": 440, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.005181238986551762, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "35baf826-1c46-56db-a032-0e97cc76924a"}}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7751995cb5b9fbf8fafe4ec7fb51da0fb18c1d1cfd2c9faa075bb48eaf3a994b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5dd928e2d9e85109b53320ff46bb9c5c9700121862c43b46d506d001fe64ca04.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7ac0d26e282cd34afaf272ae7ae512f07c41ffcdcc57710130808c7e5b883361.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8afd77630b6e854adf4567af14d3a5b00baa014baddf89b587821ca0cd895cef.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "22590027-953c-5b6a-b2ca-508afff0b457", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018935003317892551, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22590027-953c-5b6a-b2ca-508afff0b457"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5372484093531966, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30552644236013293, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8d8ca18ba0307591bae6d6fc667e8672dbee0defe7d930121606b0ab8c3ead98.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5372484093531966, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b4c92f2ff6792f76b3d3cf6dec8d14c2f8a488e0834f9642a24f0f6b67e616be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.30552644236013293, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 447, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f95d9ed17c639402c86d27fa8c6689ae0c77953699d42251e139324648a09f9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/419ba765f9c0892cd0f84aa2f72e8bf26bc0faf25efb28807da16da3cd0ff135.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c328bfa734b290f1b4eb9308df7b36e60d09221fcfd313d5b8dff8254bc121a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989827], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "f9f27f44-c195-594c-ba33-277c731bc996", "validation": {"axes": {"B": 1, "N4": 447, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004752210807055235, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f9f27f44-c195-594c-ba33-277c731bc996"}}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9c2d6fb4390c45690358c829016c5b70eb2f2ed17c3ab281a0a098a0e2726ee9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b0d280c7576b9f954e8fcd448c036b24e6450ee531d155ede8e9e06f364e5e90.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/0524c704a60b543ffa7575912a6dbcfd09bd0c50dc80e474a7987514bc45e0dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/1c6ce462fa6d7422ae594bd5e796548deeac649395162b0fb6154976a6613c98.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "6dd20c52-b8f7-5c0a-9a19-40c3e9f5c8d1", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0018482441082596779, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6dd20c52-b8f7-5c0a-9a19-40c3e9f5c8d1"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5389163577929139, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [7.450580596923828e-09], "seconds": 0.3053081319667399, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0630a5b8fe5b05b1e966f162b9af9ad4deab43a79637a4e352db98917c34bf20.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5389163577929139, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/88b3cfc2aaa50bb57cd85307bffeb877a6964104c5dfb13e7104162f68cd43c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [7.450580596923828e-09], "seconds": 0.3053081319667399, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 402, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c96a79c7877a3402cd92e63fff2c5814e620805d17eced34aae1d97dd42a2bbf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7e43282fa8e9a369fae67d91337a6a5e2a8b4d4bab2cbbf856055da844dff505.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c328bfa734b290f1b4eb9308df7b36e60d09221fcfd313d5b8dff8254bc121a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [989828], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "a13e6ed2-56fb-5d60-97ae-1abffcfe4ce1", "validation": {"axes": {"B": 1, "N4": 402, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.004425243008881807, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a13e6ed2-56fb-5d60-97ae-1abffcfe4ce1"}}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9363d93c51932542601955d0e76ea3af6382b6534d3576e5e2d4db8abf149d53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/e032ffbbf4230e554361cb952e5006e5718589bf65ff403cbdbdb71b210b1992.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7b40b0acfab32497253315212a3dd11cfe314eb7f53b26bd27b516dd8102c470.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/87f2b7e6239721531e6b775ffaed1c9dfdf3c9102020db0495a11033fcbddac2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 601, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "1b7674e8-c16e-54fc-98be-3fc37691c059", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.0018679341301321983, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1b7674e8-c16e-54fc-98be-3fc37691c059"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "selection_step": 601, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.6982219400815666, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "selection_step": 601, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3068130211904645, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f9e230d898ba149d5c73c09a3ead21604daa0f69eb4071799ae532c6c4b33481.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "selection_step": 601, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.6982219400815666, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/af2c27bbdb367588002852e0c04e855500077bc9b5f6e5a7080fb1e8bbf399e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "selection_step": 601, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3068130211904645, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 370, "N8": 1, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/a345afa325b0fb8176f03830dc1c14413bee3ecbec38e459b824180d90295ca1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/89723945dfacec4354824e18c80b1873888eafd5b50ec9334cc1dee82d76b1fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c328bfa734b290f1b4eb9308df7b36e60d09221fcfd313d5b8dff8254bc121a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [990426], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 601, "uuid": "60d85ebd-5bc6-5fb7-8062-e4646777caea", "validation": {"axes": {"B": 1, "N4": 370, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004576727747917175, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "60d85ebd-5bc6-5fb7-8062-e4646777caea"}}
 {"actual_requested_output_tokens": 1204, "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/09992723f59f11433609b144a3ec27b3fbc094e7f6fdc520693b5b679ca1a403.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a6854895c6fef58815a8941c4594e1dee46dd79f135be4c0b1c4a44f797a497d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/473bba7a3533a50f39cb0c551f5c0752f583d0d17df2fe891d0c684c951fce3f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8782c87d4f937417902687fa53270c94a20a8e34ef32a5185afcd51dfe829eaa.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "layer_id": 0, "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1202, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "0cf25067-8587-5fe2-b8b8-3093a243e8a2", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.4901161193847656e-08], "seconds": 0.0018994337879121304, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0cf25067-8587-5fe2-b8b8-3093a243e8a2"}, "variant": "decode_1"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "selection_step": 1202, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8692195420153439, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "selection_step": 1202, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3060846603475511, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0dc6aaa1cf33dfffbdedbb27a96f472dc97d2da50a90cab5c2a183ea540130e9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "selection_step": 1202, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8692195420153439, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438"}, "variant": "decode_1_cached_False"}
+{"actual_requested_output_tokens": 1204, "adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1136ae48ecf37fd4fff834667dd5024df6acab6ce12040204301159f20399590.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "selection_step": 1202, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.3060846603475511, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793"}, "variant": "decode_1_cached_False"}
 {"actual_requested_output_tokens": 1204, "adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_profile": {"block_size": 64, "branch_id": null, "hash_id_scope": "local", "in_256k_variant": false, "input_tokens": 989824, "label": "full_context_max", "locator": "requests[1920]", "model": "claude-opus-4-8", "output_tokens": 1204, "reusable_prefix_tokens": 987584, "reusable_prefix_tokens_256k": 253440, "role": "main", "selection": {"field": "input_tokens", "note": "Near-million-token tail; stress case, not a typical request.", "percentile": 100, "target": 989824}, "session_id": "5b6e131844fbcb4a57f8cf516831dc679004", "t": 192785.654, "uncached_tokens": 2240, "uncached_tokens_256k": 736384}, "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 416, "N8": 2, "P4": 128, "P8": 256}, "cache_predecessor_locator": "requests[1919]", "capture_decode_steps": [0, 1, 2, 3, 601, 1202], "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/68b09500e07c66f35fa9eab66df0eb7d88296a2f9b71cd5a01aac064a7766d15.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "input_ids_sha256": "28f0ba8d30cb3d2adef488d08b0f1e7952d8f2f679a3eb3355ba465ab9fcf012", "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c2cf5c31d7bb98642318c0acfc69a7b8d72a60bb0bf93e8e1a3010d893b4c97b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/bac456d3b5829f27b9fa8fc4259ddb68d857def1e4ec3e030065043e5491b0bd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_full_context_max", "scheduler_mode": "decode", "scheduler_sequence_lengths": [991027], "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1202, "uuid": "e40b517a-47c2-5f8d-bb80-53962880c9e7", "validation": {"axes": {"B": 1, "N4": 416, "N8": 2, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.005160517990589142, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e40b517a-47c2-5f8d-bb80-53962880c9e7"}}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4096, "TopK": 128}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924", "validation": {"axes": {"M": 4096, "Nkv": 4096, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.27322842460125685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924"}, "variant": "rows_4096_compression_0_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4096}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/764855204f9cc3e6f07d91ad9d1c08d07bd9fbf315d2e35a55ea5face56b9e78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924", "validation": {"axes": {"M": 4096, "Nkv": 4096}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.27322842460125685, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924"}, "variant": "rows_4096_compression_0_cached_False"}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/cfa6bbd7d01286eb7e5d7c65a7008d726d7973153a4757b10425033eeaa7292e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8f583f17e0de173de6c6c4efb3d1843a5d9e14e2f1e733b1184b0eecd1fc5eb8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/c9fd4bfc697b50f85dfa06eabe24a02264bd310c0e0ae3b53525343a45cb663f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/66c3deaf8590c8e10a82c739636202fc7c4666ab9abf7ac655a24f39a3a6911b.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "1a7483d5-5abb-5604-9217-2050ffde435c", "validation": {"axes": {"M": 4096}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.2782554626464844e-06], "seconds": 0.009584414772689342, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a7483d5-5abb-5604-9217-2050ffde435c"}, "variant": "prefill_4096"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "0b1815f1-1059-590a-87ae-21879f49482d", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2033551116473973, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b1815f1-1059-590a-87ae-21879f49482d"}, "variant": "prefill_4096_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.4996452019549906, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4"}, "variant": "prefill_4096_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 6144, "TopK": 640}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5", "validation": {"axes": {"M": 4096, "Nkv": 6144, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.21970500983297825, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5"}, "variant": "rows_4096_compression_2_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 8192, "TopK": 640}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7fd83f55-ee1c-5bec-b886-047096621576", "validation": {"axes": {"M": 4096, "Nkv": 8192, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.1358597120270133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7fd83f55-ee1c-5bec-b886-047096621576"}, "variant": "rows_4096_compression_1_cached_False", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc", "validation": {"axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.22201813105493784, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc"}, "variant": "rows_4096_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.190809722058475, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0"}, "variant": "prefill_4096_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.4897424881346524, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a"}, "variant": "prefill_4096_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 8319, "TopK": 640}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402", "validation": {"axes": {"M": 4096, "Nkv": 8319, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.13585394481197, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402"}, "variant": "rows_4096_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 12415, "TopK": 640}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248", "validation": {"axes": {"M": 4096, "Nkv": 12415, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.13608363270759583, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248"}, "variant": "rows_4096_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 4286, "TopK": 128}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9", "validation": {"axes": {"M": 4032, "Nkv": 4286, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.21675625583156943, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9"}, "variant": "rows_4032_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78724af1bed4afa2eccf9a9528b1bf4a13639f7ccb76d71b997a64e8d0f88373.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "0b1815f1-1059-590a-87ae-21879f49482d", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2033551116473973, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b1815f1-1059-590a-87ae-21879f49482d"}, "variant": "prefill_4096_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f93666ace0ac315f3e7f7c898091c58a43abab7dd7f0447c99545c13e6b8ac56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.4996452019549906, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4"}, "variant": "prefill_4096_cached_False"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 6144}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d5c32961f1ec9ab5d15516cf5c24897dc1c20ddc1dd2e65dd77594433dc1dcbd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5", "validation": {"axes": {"M": 4096, "Nkv": 6144}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.21970500983297825, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5"}, "variant": "rows_4096_compression_2_cached_False"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 8192}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/99c0d5885a99174b4a3b0c4c9da7b57fbdbd07bde57786f8a39e9738972edec6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [4096], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "7fd83f55-ee1c-5bec-b886-047096621576", "validation": {"axes": {"M": 4096, "Nkv": 8192}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.1358597120270133, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7fd83f55-ee1c-5bec-b886-047096621576"}, "variant": "rows_4096_compression_1_cached_False"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 4223}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/55cd5d99e9988e623c504a4e39860ab107cc895b5189159b0adcc3e4a5665dba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc", "validation": {"axes": {"M": 4096, "Nkv": 4223}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.22201813105493784, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc"}, "variant": "rows_4096_compression_0_cached_True"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e57f0d9bfa3a180ad9c23f084e909258244c223d642e216ff9e3e64118fd7dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.190809722058475, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0"}, "variant": "prefill_4096_cached_True"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18b0875290346a2ab8cf87edf2c7e79f3a74ed66d74dd0e5e3dea0e5d69096e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a", "validation": {"axes": {"M": 24576}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.4897424881346524, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a"}, "variant": "prefill_4096_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 8319}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/e2f6fa28838a8933d169d786f4d68604a5e325245b90bb1c9c32ffb122ef08ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402", "validation": {"axes": {"M": 4096, "Nkv": 8319}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.13585394481197, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402"}, "variant": "rows_4096_compression_2_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4096, "Nkv": 12415}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/7ec3de10ef7e2994921ef173190d2e76ded5926feace029a3f002aafdaf17ecc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [8192], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248", "validation": {"axes": {"M": 4096, "Nkv": 12415}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.03125], "seconds": 0.13608363270759583, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248"}, "variant": "rows_4096_compression_1_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 4286}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/a77bf95a04fd1b3eb9e91394d5968fd63fd558730f3317a6c98ec43eafdc0ed1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9", "validation": {"axes": {"M": 4032, "Nkv": 4286}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.21675625583156943, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9"}, "variant": "rows_4032_compression_0_cached_True"}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9fa83b55031c6da0f4dd6eb39e2090c86cf8c347ebc6ec6371a5d9b6a6c107a4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/73ad4498b4d4e60062bac30e73443aca13ad6669594a01a10173b1b70fff8f61.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b1569af7a8e420dc9b2db952010a7227f25877371887f86c373026eb720bc9b4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8f322960affe94c26245d79f46be5dfae183c3769ac3c07c935d3dbb3d345ed3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "ecdd3d80-02ae-533f-a687-198443ccfdca", "validation": {"axes": {"M": 4032}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.5762786865234375e-06], "seconds": 0.0119220782071352, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "ecdd3d80-02ae-533f-a687-198443ccfdca"}, "variant": "prefill_4032"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24192, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "c24c30ba-f971-5480-9756-4906df92201e", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24192, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2006504167802632, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c24c30ba-f971-5480-9756-4906df92201e"}, "variant": "prefill_4032_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24192, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24192, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.48242560867220163, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8"}, "variant": "prefill_4032_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 39070, "TopK": 640}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5", "validation": {"axes": {"M": 4032, "Nkv": 39070, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.14000864699482918, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5"}, "variant": "rows_4032_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 73854, "TopK": 640}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045", "validation": {"axes": {"M": 4032, "Nkv": 73854, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.14596810471266508, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045"}, "variant": "rows_4032_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 2238, "TopK": 128}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f", "validation": {"axes": {"M": 1984, "Nkv": 2238, "TopK": 128}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.12401726329699159, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f"}, "variant": "rows_1984_compression_0_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k128"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24192}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/45ec6ab89391b2b94ea6f0c220b32849c43f721bca937cbc6936ea8f17271ab7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "c24c30ba-f971-5480-9756-4906df92201e", "validation": {"axes": {"M": 24192}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 1.2006504167802632, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c24c30ba-f971-5480-9756-4906df92201e"}, "variant": "prefill_4032_cached_True"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 24192}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/04885efb8cd8095b0626dc86ce33245cf9403d3c75587cada67df74ecf0116f5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8", "validation": {"axes": {"M": 24192}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.48242560867220163, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8"}, "variant": "prefill_4032_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 39070}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/2099ef68a72e44bebc6a4ea7e0f0001d2322a16834423a65594d9a445e44f9f8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5", "validation": {"axes": {"M": 4032, "Nkv": 39070}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.14000864699482918, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5"}, "variant": "rows_4032_compression_2_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 4032, "Nkv": 73854}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/c225935ec58ab02c9d3e34267901d49661d41fbeff521fb7c95756e851cecb5c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35008, 34560], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 8, "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045", "validation": {"axes": {"M": 4032, "Nkv": 73854}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.14596810471266508, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045"}, "variant": "rows_4032_compression_1_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 2238}, "compression_ratio": 0, "definition": "sparse_mla_prefill_h64_d512_k128", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/2d74e9a5b97064d349ebec93d921024706e89a2912007c7808e0af74652491c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f", "validation": {"axes": {"M": 1984, "Nkv": 2238}, "definition": "sparse_mla_prefill_h64_d512_k128", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.12401726329699159, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f"}, "variant": "rows_1984_compression_0_cached_True"}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b634502a4eb52d5e6f5578329bb2d694f92bf423d0b47d5d33e80542b8a357c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/267ec2804c72a33a306ad7c882ff27c7f688b188fe43c392d4f1efd369d694d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/3b1bd47e74b20959e285c20b5ac47a292018aef09eabad9bed688bc50f80e396.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ec13eb5bcf3dba7f9855ce0baf02107ea76c9335237b05142b4e559a5619e419.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "491c4277-463b-57a8-b65a-38cc0439366b", "validation": {"axes": {"M": 1984}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 3.2186508178710938e-06], "seconds": 0.0095808538608253, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "491c4277-463b-57a8-b65a-38cc0439366b"}, "variant": "prefill_1984"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 11904, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 11904, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.8494550948962569, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56"}, "variant": "prefill_1984_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 11904, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 11904, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.45867891516536474, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409"}, "variant": "prefill_1984_cached_True", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 37662, "TopK": 640}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797", "validation": {"axes": {"M": 1984, "Nkv": 37662, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.08471962995827198, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797"}, "variant": "rows_1984_compression_2_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
-{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 73086, "TopK": 640}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "43e73df3-0172-583e-944e-789dff99b8b6", "validation": {"axes": {"M": 1984, "Nkv": 73086, "TopK": 640}, "definition": "sparse_mla_prefill_h64_d512", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.0831243572756648, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "43e73df3-0172-583e-944e-789dff99b8b6"}, "variant": "rows_1984_compression_1_cached_True", "captured_definition": "sparse_mla_prefill_h64_d512_k640"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 11904}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/76f5455e6821e3a552ca4714431200a1cc59f16e82526871fe6226e1fa6090dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56", "validation": {"axes": {"M": 11904}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.8494550948962569, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56"}, "variant": "prefill_1984_cached_True"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 11904}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/82ec663da9cb8e1785b6e27599ee8abad1158d6200ae714513bfe17c3d55d101.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409", "validation": {"axes": {"M": 11904}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.45867891516536474, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409"}, "variant": "prefill_1984_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 37662}, "compression_ratio": 2, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/593b16763e612bc299b370621427e32bd9fc285a2e74bf1e3223cde14f51c7ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797", "validation": {"axes": {"M": 1984, "Nkv": 37662}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.08471962995827198, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797"}, "variant": "rows_1984_compression_2_cached_True"}
+{"adapter": "Gather all logical query heads and outputs from tensor-parallel ranks; retain the complete BF16 workspace and original sparse indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1984, "Nkv": 73086}, "compression_ratio": 1, "definition": "sparse_mla_prefill_h64_d512_k640", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/719f6f15791967aa4272baf8c5bf3c461f013c11f949a9d16eb478250befd9eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "sgl_kernel.flash_mla.flash_mla_sparse_fwd", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "prefill", "scheduler_sequence_lengths": [35392, 35456], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 9, "uuid": "43e73df3-0172-583e-944e-789dff99b8b6", "validation": {"axes": {"M": 1984, "Nkv": 73086}, "definition": "sparse_mla_prefill_h64_d512_k640", "device": "cuda", "max_absolute_errors": [0.015625], "seconds": 0.0831243572756648, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "43e73df3-0172-583e-944e-789dff99b8b6"}, "variant": "rows_1984_compression_1_cached_True"}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2d668b2bdde0e5a33414874ad0837a906621f9e319fb4ee5a6ea4f6cdfc4711a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ad002998d952e2da467a96a7f9512e4426c249e0ffec6764bf79aac23d572001.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/70607226c53c2d7251994f07fd4ea48fce00b8309715f0d3f68d5bef1de7a86d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/a446ce78deef3c4deb3c519281a19b6fb7897115d9612d2644dfb155992c36fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "8fb5668d-9ace-5884-ae82-5cfb3e534ef0", "validation": {"axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 4.470348358154297e-08], "seconds": 0.003166140988469124, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8fb5668d-9ace-5884-ae82-5cfb3e534ef0"}, "variant": "decode_3"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [3.0517578125e-05], "seconds": 0.5468553341925144, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [1.4901161193847656e-08], "seconds": 0.31446270225569606, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9d4842fc59ce46e337942b1cba771d037ff740095aa7142979d49b7aae1029ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [3.0517578125e-05], "seconds": 0.5468553341925144, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694"}, "variant": "decode_3_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c9a161ba440d995c1fef182947f196952c92f8a1526074b4c4496a7836f1ee5a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 0, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [1.4901161193847656e-08], "seconds": 0.31446270225569606, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5"}, "variant": "decode_3_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 3, "N4": 146, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f5df05e0ec51a60f36ca4e451427d94c1f29a31b08b3c08f194600ceb579b7f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/a2896a2c7a77e677a03469000528f09853cb68ca8f38829a16dbe1df13ca2527.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c9f52977c1c53b43c87c856fd4e33082f009452798662f3300c78b31a1ae9245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "506de298-8285-52bb-87c1-15d03a7a50d1", "validation": {"axes": {"B": 3, "N4": 146, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.006262518931180239, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "506de298-8285-52bb-87c1-15d03a7a50d1"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/39a8b50e8b4fc730544481896881e88b41a921174982aae68b63bd2f1b6ef0c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9a4526719c278596443909500b38e29f7d50bbc8cd648b825b9e783f57313aff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/15e101bc4551272a345ef15e7c05f2bbb66e2705bfc58c2be631ad5245f8256b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/abc5eb2e8216456dfcad63b2ee8ddc72080be9456a0539f8856ff68ef505a0ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "b22e0ab8-07f2-5028-b601-d5e2771d3d2c", "validation": {"axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.003138778731226921, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b22e0ab8-07f2-5028-b601-d5e2771d3d2c"}, "variant": "decode_3"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.5430445061065257, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "22791a4a-6151-5445-806d-c3793c972208", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.31428917311131954, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22791a4a-6151-5445-806d-c3793c972208"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ba220fbeebfbd704489d0a0787e0cd8693745eae6b39d0c2ac80f628708bd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.001953125], "seconds": 0.5430445061065257, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f"}, "variant": "decode_3_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e27eb4ceb7a5548020a18cb0a1588d8b56f2419bdabdea8cf58afa6fa22cb0cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 1, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "22791a4a-6151-5445-806d-c3793c972208", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.00048828125], "seconds": 0.31428917311131954, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "22791a4a-6151-5445-806d-c3793c972208"}, "variant": "decode_3_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 3, "N4": 147, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/82be75d76af79c535b3246ec7a5b9266b1797e6d83adf6b67b6ab143f136b947.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/8dac3a9477bb14818a835bc00f9c8d7e25000c1808760c2a1c989d690bba8818.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c9f52977c1c53b43c87c856fd4e33082f009452798662f3300c78b31a1ae9245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35010, 35394, 35458], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 1, "uuid": "280fce72-38f0-5ddd-9ce7-ed847babecfc", "validation": {"axes": {"B": 3, "N4": 147, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004787764046341181, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "280fce72-38f0-5ddd-9ce7-ed847babecfc"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/80585fbbd2d8cdbf3deccb7fee2ca53482729bccac0be1297ffb561a0e93f73b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f8674e58b6644b08d75d019456aad2d848727ba7ee540889e181818e7139451f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/eb188f2e6a71a78009018322d159f5cc600e84faac6672bc00c9f8fd0b8679ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/19405bc4c99b9d5cb7b352b177a855747c7d6a7ca93f9d29324076ec696ed8da.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "c25f65ed-4124-5afa-b47d-07d571b319eb", "validation": {"axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.015995109919458628, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c25f65ed-4124-5afa-b47d-07d571b319eb"}, "variant": "decode_3"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5427321679890156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.47860030410811305, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7341d2ea5331fa1ffccd2b88437fa8179fa288eed5331e0053d1e4b20132799b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5427321679890156, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831"}, "variant": "decode_3_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c0c963830f7d18156b7a27526c487f0d00a1dd6d146cfd57b05a18a61b048e99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 2, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.47860030410811305, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a"}, "variant": "decode_3_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 3, "N4": 145, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/cebc0f71b12521cca5bf67fce10baeba1bcc98fd2e1b27133a688dfde116b3f8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2f19143111599a342f90237ebbbec640fd7ee106a6c7c37c90d1f21bb3b67dab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c9f52977c1c53b43c87c856fd4e33082f009452798662f3300c78b31a1ae9245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35011, 35395, 35459], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 2, "uuid": "9e1eb156-12ab-5687-b99e-5694bc8e11f9", "validation": {"axes": {"B": 3, "N4": 145, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004719646647572517, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9e1eb156-12ab-5687-b99e-5694bc8e11f9"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/9cf39b82297378f66b74bb51c14c116a4fe474b8121cfec3cad4e2e39f6d94ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/c4087880fc82c8f309977295103024aa2f291cd4c23df6d821ed8ccfa924397b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5ced3d55cc98d8df048993f5dba53f0886e71859460bbcb48bc7c64356e76f09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/5bfdc7caeb5ed03f6bb269e131ee22b3bda2320742229e6eb1c94de77b5b823e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "bf11c6be-1cf2-5593-8167-3d13642ce599", "validation": {"axes": {"M": 3}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0033489898778498173, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bf11c6be-1cf2-5593-8167-3d13642ce599"}, "variant": "decode_3"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5432731551118195, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5164944920688868, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d"}, "variant": "decode_3_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc843cbb7c7af673836e7f79d508376468f318df69759e78653a5e3dbc6d0165.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5432731551118195, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965"}, "variant": "decode_3_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05cd5b433dcc242dfc03139f3eba7e832eb23d5289fc4770b0b4435c6ea89fc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 3, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d", "validation": {"axes": {"M": 18}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5164944920688868, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d"}, "variant": "decode_3_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 3, "N4": 145, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/f1a17d4804206a6ca6fffe747dedcd6031e3760933e0eea8f6515c0446bf974e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2f19143111599a342f90237ebbbec640fd7ee106a6c7c37c90d1f21bb3b67dab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c9f52977c1c53b43c87c856fd4e33082f009452798662f3300c78b31a1ae9245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35012, 35396, 35460], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 3, "uuid": "c4bb3ecc-abd1-544d-933e-5dfe11ca7428", "validation": {"axes": {"B": 3, "N4": 145, "N8": 4, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.004712217953056097, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "c4bb3ecc-abd1-544d-933e-5dfe11ca7428"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/4514ca9b389e49bdc0ec882f4d49de1773fdbd71eb196a0194873eff48846acb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/3e47a783f52602828428b6c3aa646e733a4cf5c3bb78f9a6d430178c59f24c96.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/38b1fbf2d0d656032280600042baca8806ddb783b11860ae3ff6c6bb964aa553.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/92a9d60c184a26b04fd4bba0ce7c80ef32c997741527ff1e8870b7003135d185.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 107, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "fa53421f-e2db-580b-8e24-ff234a18d7a9", "validation": {"axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.003359489608556032, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fa53421f-e2db-580b-8e24-ff234a18d7a9"}, "variant": "decode_2"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 107, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5384206641465425, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 107, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5091732949949801, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9aa9f5547e4f207fc0d859025d8779b2fc61cc3ae8bc068c09b6858f547fe3a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 107, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5384206641465425, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551"}, "variant": "decode_2_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d3b34d5fa8f1e8653b055649159a24526cea9acb5b8e1a471b288e1cedbfb10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 107, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5091732949949801, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577"}, "variant": "decode_2_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 2, "N4": 135, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/9b98d9134538d7ed172ca951a5997176074af8f11e26b355a8ad8fd9b8b02f51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/e9f99b6f57813b574a27f40aeda2b0ea6182df8eacf37cc74abcf8300b67e33d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7175e91a4f69f8fbae97e298a50b1e0fdd3c8cb42121c0b304f05a4db79f33b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35116, 35564], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 107, "uuid": "193a3358-bd3f-5f02-bd91-33e45717225e", "validation": {"axes": {"B": 2, "N4": 135, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0038592899218201637, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "193a3358-bd3f-5f02-bd91-33e45717225e"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b0931ec9a31a05d20e3024ffafabeefbac295d62eee347ec95ee3f9fe57a1256.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/90f02c869fa8f5456ca9ed87f62cc4e755212f1b0126fa59c90c56475a2af94f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/81277ea0c24f35dac23469d46496c069612cbb669ff0f3c9deda99c6804be28d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/ad64fa161fa575df2e029246f649b0e467687820c41e36f4f32c1a535ead015d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 108, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "82dfd151-92c5-5eff-8342-19b4e342569f", "validation": {"axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 8.940696716308594e-08], "seconds": 0.0030926777981221676, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "82dfd151-92c5-5eff-8342-19b4e342569f"}, "variant": "decode_2"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 108, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [9.5367431640625e-07], "seconds": 0.5360788740217686, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 108, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5105185238644481, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/25ca1d7e28d2c6cba057dffde7fe560203d0e5120467cf76a5cf361988c294df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 108, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [9.5367431640625e-07], "seconds": 0.5360788740217686, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12"}, "variant": "decode_2_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c90a693e662936f2e4593b1e08a01dc9726cc6751aaee18a53b47201a98d8b55.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 108, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5105185238644481, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044"}, "variant": "decode_2_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 2, "N4": 142, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/eef89e0c51aee47176fd9da00ba46ee659fab25e919b56d9e7d99809164e0196.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/3ab1b2be5aa477011182d95d4983647dbf790bf20f53898ab38e964b9b7c1f33.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7175e91a4f69f8fbae97e298a50b1e0fdd3c8cb42121c0b304f05a4db79f33b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35117, 35565], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 108, "uuid": "12180b5b-0be9-5de0-a091-1bb8e20af73f", "validation": {"axes": {"B": 2, "N4": 142, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003872013185173273, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "12180b5b-0be9-5de0-a091-1bb8e20af73f"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/472c45c2f00070e7566940168061d24f47d66eda3f7779331289ad933fdf4a88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7356a566ff0dff2ee6fd902b3fb59220abc2de2ef3e699e4377e9e0bfa3d8d23.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/65163185b87a691c8bcbb1c1f6601326f8bb79d7a5bd27302e3cf54e7ee87c3c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/fc0bb4c3b4d5eb744961c8185b55d4f956d89f53d3496afa8fa73bb596a3541b.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 109, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "4082bf7f-299b-5d20-bf12-133e52e792ab", "validation": {"axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.003034201916307211, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4082bf7f-299b-5d20-bf12-133e52e792ab"}, "variant": "decode_2"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 109, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [1.4901161193847656e-08], "seconds": 0.53823347017169, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 109, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5099127921275795, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/564affa818afc21e5f7b864eca737d89e1a78824f63d9d08feb4256f3133a49d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 109, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [1.4901161193847656e-08], "seconds": 0.53823347017169, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81"}, "variant": "decode_2_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98e010945f80933ba933dd6d90d79797886596a03e6f50ec17a7b1bf970b13dc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 109, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5099127921275795, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5"}, "variant": "decode_2_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 2, "N4": 142, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c2718b594ea5f338f2a08db9a8fa70031fdafddd79b79cb6688cc0149113e78e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/2b2069cb79c87ef945fff06ec4c6e5bd5820f1153b05f4616f13a73a371ba410.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7175e91a4f69f8fbae97e298a50b1e0fdd3c8cb42121c0b304f05a4db79f33b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35118, 35566], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 109, "uuid": "35704cb4-9369-54f7-9452-896d00e9f9d6", "validation": {"axes": {"B": 2, "N4": 142, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.003853880800306797, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "35704cb4-9369-54f7-9452-896d00e9f9d6"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/b75c48785f6884ee69a0d69e907e741bcfa70ae9aead73ef7e5391dc5765dcc0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/290a3f5b4e61b941e132d9143506f5fbf714be34318acf88058c19b294873f60.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6ef1b7fae7435e3175209b0c9cd75de8b4d862f3054d3ca151763fc0a88b8d4c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6fe15335ee2427356d678dea6d0d4530917735f05ecf27c4aa31df966f8e99b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 110, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "99713957-37e9-559a-92d5-a800373b3e75", "validation": {"axes": {"M": 2}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0032311580143868923, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "99713957-37e9-559a-92d5-a800373b3e75"}, "variant": "decode_2"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 110, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.5387282078154385, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 110, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5106046716682613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41"}, "variant": "decode_2_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7171098400c11f313a1f99a27d1b25d64ee3b9bf0e862472f9b6de4d35eb9d26.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 110, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0001220703125], "seconds": 0.5387282078154385, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c"}, "variant": "decode_2_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/289bcc3be71bdaf8abb4dd561dbdb7612c4b546a483e6c83116a4e6ca6f2ae71.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 110, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41", "validation": {"axes": {"M": 12}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5106046716682613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41"}, "variant": "decode_2_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 2, "N4": 135, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/b396ef2bc6965cfd6bd24b241b84110341cf817221e89c196d9276ed9636b489.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/bc1633e87b1e444fc238af8089f6989762b95d63a2fa9fefae3203ca1035f1a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7175e91a4f69f8fbae97e298a50b1e0fdd3c8cb42121c0b304f05a4db79f33b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35119, 35567], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 110, "uuid": "b74cdd36-78e7-5612-ad84-cc44b15c5b29", "validation": {"axes": {"B": 2, "N4": 135, "N8": 3, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.00390625], "seconds": 0.00393477501347661, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "b74cdd36-78e7-5612-ad84-cc44b15c5b29"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/380c748ec6377c69e4f4dae5bdd8ad09cdb1446b65483d1ac54eaa5ed9c20963.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/0b46ff7dcbf13b37156459f61cfc5a43e0c825f0abaf8aa98f860fb2c04abba8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/1c2b58efbd07006dd8346ed6c32c17893192685781f7c5f84dc27f47032acf8e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/2fae900b61c5c2f6e097d6a9efb4e2b8d3545cf6646138e166d9e3ec930ac853.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 119, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "eea430f9-5bcb-58a9-b079-c1bd8f062f3b", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 1.1920928955078125e-07], "seconds": 0.003239080309867859, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "eea430f9-5bcb-58a9-b079-c1bd8f062f3b"}, "variant": "decode_1"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 119, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5724243517033756, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 119, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5190592082217336, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0aa68cce6b21285e516bb732cbbfd2fb48834a4d2a70dcddc159d1cb8039024a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 119, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5724243517033756, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869"}, "variant": "decode_1_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/225488cfaa707501dc60747f67d8baefe00d1536d12bf0cb329c999018ab691f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 119, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5190592082217336, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc"}, "variant": "decode_1_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 122, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/79ec5b681bdaf2575efced183ff5b8fb657722eaab94c4e6080a6bbf12d729c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/e0c9aeb5080f443ef65a4ade56b5d1ff3007998177e9468237fabde133ec4755.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d2fa94370cd62981deb374c2e9d7dff2abdb59ac751a6da4a55b2514f87fd6de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35576], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 119, "uuid": "0f727df7-b91d-5b0e-a1d7-0a24dc243af6", "validation": {"axes": {"B": 1, "N4": 122, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.002870344091206789, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "0f727df7-b91d-5b0e-a1d7-0a24dc243af6"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f3e16897e5635b06ad236a9b0159273b54386b9d2d4b728ddb5fbf31d40f7893.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/7609f7c926b42d80cc4f7a2b942bdafd6b7a4d932561756a9a11fbab37c821ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/c5fc2c52ac0c28f18030c03f884ca300f1632e20208d62c300200af562b711c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/f0788eae3193a3437bd228993cf5e0cf56c91cd7bec90c007f333b42a6f2c02c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 120, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "51a3452f-7f5d-5c24-80a6-8c2f7f9bcf23", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 2.9802322387695312e-08], "seconds": 0.003138733096420765, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "51a3452f-7f5d-5c24-80a6-8c2f7f9bcf23"}, "variant": "decode_1"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 120, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8617571918293834, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 120, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5206519616767764, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f956d858cf6f80a86747eb7a5795eec2a713e598036d9ae80d4c96c8c119983.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 120, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8617571918293834, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8"}, "variant": "decode_1_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0871c01e37b7892d7b53ef86b9c2d10365f801a77d2922b2a7f5c57a82737c9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 120, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5206519616767764, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8"}, "variant": "decode_1_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 129, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7a80eb686d7035011defb2775d32ad9566567e41b2d45c65e852af4299f7bf53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/4c8da3f45f8baed076a2e521bb8fba92e91a8b5dd7edebcfc4a8f995d4f7ba23.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d2fa94370cd62981deb374c2e9d7dff2abdb59ac751a6da4a55b2514f87fd6de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35577], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 120, "uuid": "9d1297e1-e7b8-5427-9dbd-e166c59fe58e", "validation": {"axes": {"B": 1, "N4": 129, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0028816862031817436, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "9d1297e1-e7b8-5427-9dbd-e166c59fe58e"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6d60916011cde578ebcccad4448e657c00c49f53c0a4361a4532f739385f84a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/75235ab00e75190806964cd429e4b81989d9a6862ce320e2f725d25a4fb11240.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/db56d04b702ccbf75b96ada555e5abb652f0d3b5250528cea10a76a0bc5b3995.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/6ed2fdca2277deb9b7778bb6d055d80bc2b65c68794701f3f45b9fab9ea5d976.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 121, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "4c2da6e9-5a34-5b15-9b7b-689d4389f09d", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 4.470348358154297e-08], "seconds": 0.002981953788548708, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "4c2da6e9-5a34-5b15-9b7b-689d4389f09d"}, "variant": "decode_1"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 121, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8776261690072715, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 121, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "7713b508-dfc6-502f-be67-177b21b7c674", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5179842039942741, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7713b508-dfc6-502f-be67-177b21b7c674"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f5bd17366a02161d6946d489d3bce39189c8b5410978cd1e093801f90ac08e7b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 121, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8776261690072715, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae"}, "variant": "decode_1_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/83b729789ab6435d41414a6e6ca79af5848a6ed331198457457d0a460f72b265.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 121, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "7713b508-dfc6-502f-be67-177b21b7c674", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.5179842039942741, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "7713b508-dfc6-502f-be67-177b21b7c674"}, "variant": "decode_1_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 123, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/7e40c73d64119f7df96fd61dfbe3fbc7fe638280f7eeea5f49df9e6033757489.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c6c1e11fb47edd5b81dd10f3fa491a5a4df1adb5123ec582b830fd742b715144.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d2fa94370cd62981deb374c2e9d7dff2abdb59ac751a6da4a55b2514f87fd6de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35578], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 121, "uuid": "61334221-7f5b-5c6f-8650-664972cff6d1", "validation": {"axes": {"B": 1, "N4": 123, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0028644781559705734, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "61334221-7f5b-5c6f-8650-664972cff6d1"}}
 {"agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/09ccd29f0829f65025190c659895407160ecff9b3acd63ea211290aba52e59eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/13c28eed277a2c6df1df33376c37b8833286cf7318a269ae7e2ffbb56d61975f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"expert_ids": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/21c20f4cd20758f90ac00e0362996652d00ef5c67ad1f3fb19c4a96a72f52c94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_weights": {"path": "blob/workloads/moe_router_gemm_topk_deepseek41_d5120_e384_topk6/8b9dcb083ab7313fb9f1781b10bd9b9a891b0298f317af4dd05a3ef89963b9ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 0, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "MoEGate.forward followed by vision_topk", "output_adapter": "Apply the engine's deferred routing scale to its selected weights.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_order": "Serialize the unchanged selected expert set by increasing expert ID; permute the actual weights with their IDs.", "selection_step": 122, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "04b439a3-46d3-5070-9794-a00f75da3524", "validation": {"axes": {"M": 1}, "definition": "moe_router_gemm_topk_deepseek41_d5120_e384_topk6", "device": "cuda", "max_absolute_errors": [0, 5.960464477539063e-08], "seconds": 0.0030375015921890736, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "04b439a3-46d3-5070-9794-a00f75da3524"}, "variant": "decode_1"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 122, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0", "validation": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8652677489444613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32"}
-{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 122, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "94998979-421b-5dc0-a611-c2abe525b98d", "validation": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "definition": "grouped_gemm_fp8_fp4_e384_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.526547497138381, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "94998979-421b-5dc0-a611-c2abe525b98d"}, "variant": "decode_1_cached_False", "captured_definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/628150af8edcc16649d3f5c05d8b336d82cffd7a93d17e918f63b08595aabbab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 122, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.8652677489444613, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0"}, "variant": "decode_1_cached_False"}
+{"adapter": "Gather all expert shards; discard rows marked -1 by the engine; preserve group-32 activation scales or expand each actual group-128 scale into four identical bytes.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b3da1cf1cdb650bcfebd794569e2eaebef20f4f4fab374932924902f728d7803.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "SGLang DeepGEMM grouped GEMM", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "selection_step": 122, "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "94998979-421b-5dc0-a611-c2abe525b98d", "validation": {"axes": {"M": 6}, "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "device": "cuda", "max_absolute_errors": [0.0], "seconds": 0.526547497138381, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "94998979-421b-5dc0-a611-c2abe525b98d"}, "variant": "decode_1_cached_False"}
 {"adapter": "Gather real query heads from all tensor-parallel ranks; preserve all referenced physical pages and remap indices.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_4": [422126, 128, 1, 288], "allocated_cache_8": [234, 256, 1, 528], "axes": {"B": 1, "N4": 125, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "engine": "SGLang", "engine_adapter": "Pass actual dispatcher activation-group metadata to DeepGEMM; the pinned FP4-expert recipe otherwise hardcodes group 128 for group-32 activations. Tensor bytes and model architecture are unchanged.", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/680198fe3771d2425a4a5a9ef571afe84bd7651769606315499ea1cfa1028d5f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "deepseek-ai/DeepSeek-V4.1-Flash", "model_revision": "5254448103b6254fc646dacbca1cbbf20e9216c1", "operation": "flash_mla_with_kvcache mixed FP8/FP4", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "physical_pages_4": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/c84e6b21d4b3fc07c70dd0d693f32713087c292b3c08f67c71a412765402056d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "physical_pages_8": {"path": "blob/workloads/mixed_mla_decode_deepseek41_h64_d512_k8128_k4512/d2fa94370cd62981deb374c2e9d7dff2abdb59ac751a6da4a55b2514f87fd6de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "deepseek41_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35579], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/deepseek41", "--host", "127.0.0.1", "--port", "30100", "--dtype", "bfloat16", "--trust-remote-code", "--tp-size", "4", "--ep-size", "4", "--moe-runner-backend", "deep_gemm", "--moe-a2a-backend", "none", "--disable-shared-experts-fusion", "--disable-cuda-graph", "--disable-overlap-schedule", "--cuda-graph-backend-prefill", "disabled", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.9", "--max-running-requests", "24", "--watchdog-timeout", "3600"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_DSV41_ENGRAM_HOST_TABLE_LAYOUT": "shared", "SGLANG_DSV4_COMPRESSED_KV_LAYOUT": "auto", "SGLANG_DSV4_KV_LAYOUT": "v41", "SGLANG_ENABLE_DSV41_ENGRAM_HOST_TABLE": "1", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 122, "uuid": "74f468ba-1533-57c6-af5d-746c40d970a9", "validation": {"axes": {"B": 1, "N4": 125, "N8": 1, "P4": 128, "P8": 256}, "definition": "mixed_mla_decode_deepseek41_h64_d512_k8128_k4512", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.0029348651878535748, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "74f468ba-1533-57c6-af5d-746c40d970a9"}}
 {"adapter": "Retain all live pages byte-for-byte and remap their page IDs; omit unreachable allocator pages.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 16, 1, 128], "axes": {"B": 3, "MP": 554, "NP": 580, "P": 16}, "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "engine": "SGLang", "engine_block_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/4cf63375da0395c5e3ca60c230717e9871f5f29e75da12b0c26419ffc8f53224.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/73ad33758cdcfa87daec087ebfcfc107c69b20cbf67b700d9e25a530b24c0b90.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 3, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QSAIndexer.select_decode_tokens", "original_physical_page_ids": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/2c1d36eab8b431730d31aa1bef10ce7dc5fd354f16c63f2fcc4985f99d53d41e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "output_adapter": "Serialize the engine's selected set in increasing logical-ID order; preserve membership and -1 suffix padding.", "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "padded_query": {"path": "blob/workloads/qsa_index_topk_decode_qwen38_h4_d128_topk512_c4/13f5956ada05905617cb6203847f0421597fb4a7f7991cdefc713450e82e205d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "query_adapter": "Remove verified zero query heads inserted for the backend's eight-head tile.", "request_id": "qwen_next_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "5ec0101c-535c-580f-992e-53ed104e883c", "validation": {"axes": {"B": 3, "MP": 554, "NP": 580, "P": 16}, "definition": "qsa_index_topk_decode_qwen38_h4_d128_topk512_c4", "device": "cuda", "max_absolute_errors": [0], "seconds": 2.7295097089372575, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "5ec0101c-535c-580f-992e-53ed104e883c"}}
 {"adapter": "Snapshot after the current-token KV write; retain every live page and remap physical IDs.", "agentx_dataset": "semianalysisai/cc-traces-weka-062126", "agentx_revision": "23f152f6f0f9399a85901b89a6458def0ef16729", "allocated_cache_shape": [83743, 64, 2, 256], "axes": {"B": 3, "MP": 555, "NP": 583, "P": 64}, "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "engine": "SGLang", "engine_version": "0.0.0.dev1+g684fea8af", "expected_output": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/fa7a8de18c6b22bf96316fb9b0181e73b8da871a0ad38a3d4e5b515151dab9ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "fixture_sha256": "910fedaa22efdfa0ef4df0b2f2d13385622db054c30f1cc18438b300a7f9aba4", "layer_id": 3, "members": [{"block_size": 64, "branch_id": "subagent_002_7bd2099b", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35008, "locator": "requests[49].requests[0]", "model": "claude-haiku-4-5-20251001", "output_tokens": 120, "reusable_prefix_tokens": 34368, "reusable_prefix_tokens_256k": 34368, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86543.763, "uncached_tokens": 640, "uncached_tokens_256k": 640}, {"block_size": 64, "branch_id": "subagent_003_788ad06a", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35392, "locator": "requests[50].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 108, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86549.019, "uncached_tokens": 384, "uncached_tokens_256k": 384}, {"block_size": 64, "branch_id": "subagent_001_f552fe6f", "hash_id_scope": "local", "in_256k_variant": true, "input_tokens": 35456, "locator": "requests[48].requests[1]", "model": "claude-haiku-4-5-20251001", "output_tokens": 126, "reusable_prefix_tokens": 35008, "reusable_prefix_tokens_256k": 35008, "role": "subagent", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207, "uncached_tokens": 448, "uncached_tokens_256k": 448}], "model": "nvidia/Qwen3.8-Flash-Next-NVFP4", "model_revision": "fc694b54fb0174e0913e6adf86691ef85a4ead47", "operation": "QwenSparseAttnBackend.forward_decode", "original_physical_page_ids": {"path": "blob/workloads/qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4/1fee580281d8edb69d210d9cd5eda1ab1d6659fd1668040c0de3ea620cff6c55.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "overlap_profile": {"batch_size": 3, "label": "overlap_p50", "model": "claude-haiku-4-5-20251001", "session_id": "002001296e8a8c38ad9d7cc436d691afc602", "t": 86550.207}, "prompt_reconstruction": "Public AgentX hash blocks select deterministic spans of an authored code corpus; original prompt text is unavailable. Input IDs preserve exact recorded block lengths.", "request_id": "qwen_next_overlap_p50", "scheduler_mode": "decode", "scheduler_sequence_lengths": [35009, 35393, 35457], "scheduling": "Submit the observed overlapping requests together; record the actual engine batch, without claiming to reproduce their original timing.", "server_command": ["python", "-m", "sglang.launch_server", "--model-path", "/task/models/qwen-next-nvfp4", "--host", "127.0.0.1", "--port", "30200", "--dtype", "bfloat16", "--moe-runner-backend", "flashinfer_cutlass", "--ple-offload-embedding", "--linear-attn-decode-backend", "triton", "--linear-attn-prefill-backend", "triton", "--mamba-ssm-dtype", "float32", "--disable-cuda-graph", "--disable-overlap-schedule", "--chunked-prefill-size", "4096", "--mem-fraction-static", "0.8", "--max-running-requests", "24", "--max-mamba-cache-size", "120"], "serving_environment": {"SGLANG_BUILD_COMMIT": "684fea8af3b11a5b51ec9966c02959a32ad3421c", "SGLANG_BUILD_URL": "https://github.com/sgl-project/sglang/actions/runs/38489800016", "SGLANG_CACHE_DIR": "/task/cache/sglang", "SGLANG_IMAGE_TAG": "lmsysorg/sglang:nightly-dev-20261014-684fea8a", "SGLANG_RUST_BUILD_MODE": "never"}, "step": 0, "uuid": "d8db059c-8442-54c9-a179-9c32a5bfd784", "validation": {"axes": {"B": 3, "MP": 555, "NP": 583, "P": 64}, "definition": "qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4", "device": "cuda", "max_absolute_errors": [0.0078125], "seconds": 0.3524620537646115, "status": "passed", "tolerances": {"bfloat16": {"atol": 0.005, "rtol": 0.016}, "float32": {"atol": 2e-05, "rtol": 0.0001}, "gdn_prefill_state": {"max_head_relative_rms": 0.02, "reason": "The engine's chunk algorithm rounds intermediate matrix operands to BF16 although its final state is stored in FP32.", "relative_max": 0.015, "relative_rms": 0.01}, "integer": "exact"}, "torch": "2.14.1+cu130", "uuid": "d8db059c-8442-54c9-a179-9c32a5bfd784"}}
diff --git a/definitions/dense_attention/noncausal_dense_attention_h32.json b/definitions/dense_attention/noncausal_dense_attention_h32_d128.json
similarity index 91%
rename from definitions/dense_attention/noncausal_dense_attention_h32.json
rename to definitions/dense_attention/noncausal_dense_attention_h32_d128.json
index 3ffe9f347f031ef3492148ad0ff5f0072e18e20d..398bc495d5f0786239665f47530e54d4530bef38 100644
--- a/definitions/dense_attention/noncausal_dense_attention_h32.json
+++ b/definitions/dense_attention/noncausal_dense_attention_h32_d128.json
@@ -1,5 +1,5 @@
 {
-  "name": "noncausal_dense_attention_h32",
+  "name": "noncausal_dense_attention_h32_d128",
   "op_type": "dense_attention",
   "description": "Noncausal self/cross attention with FP32 accumulation and BF16 output. Self attention uses Nq=Nk; cross attention allows unequal lengths. No dropout. Projection, normalization, rotary embeddings and output gates are outside the definition. No mask: every captured call's expanded visibility mask was all True (on one GPU the engine passes no mask, or an all-visible prompt mask), so KDA drops the dataset's mask input.",
   "tags": [
@@ -25,8 +25,8 @@
       "value": 32
     },
     "D": {
-      "type": "var",
-      "description": "Head dimension: 64 for audio, 128 for video."
+      "type": "const",
+      "value": 128
     }
   },
   "inputs": {
@@ -60,7 +60,7 @@
     "sm_scale": {
       "shape": null,
       "dtype": "float32",
-      "description": "Model default 1/sqrt(D)."
+      "description": "Model default 1/sqrt(128)."
     }
   },
   "outputs": {
diff --git a/definitions/dense_attention/noncausal_dense_attention_h32_d64.json b/definitions/dense_attention/noncausal_dense_attention_h32_d64.json
new file mode 100644
index 0000000000000000000000000000000000000000..40bf10ce999a71a4a471f5c80b764c70afadc778
--- /dev/null
+++ b/definitions/dense_attention/noncausal_dense_attention_h32_d64.json
@@ -0,0 +1,83 @@
+{
+  "name": "noncausal_dense_attention_h32_d64",
+  "op_type": "dense_attention",
+  "description": "Noncausal self/cross attention with FP32 accumulation and BF16 output. Self attention uses Nq=Nk; cross attention allows unequal lengths. No dropout. Projection, normalization, rotary embeddings and output gates are outside the definition. No mask: every captured call's expanded visibility mask was all True (on one GPU the engine passes no mask, or an all-visible prompt mask), so KDA drops the dataset's mask input.",
+  "tags": [
+    "model:ltx-2.5",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "B": {
+      "type": "var",
+      "description": "Runtime batch size."
+    },
+    "Nq": {
+      "type": "var",
+      "description": "Runtime query length."
+    },
+    "Nk": {
+      "type": "var",
+      "description": "Runtime key/value length."
+    },
+    "H": {
+      "type": "const",
+      "value": 32
+    },
+    "D": {
+      "type": "const",
+      "value": 64
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "B",
+        "Nq",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "K": {
+      "shape": [
+        "B",
+        "Nk",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "V": {
+      "shape": [
+        "B",
+        "Nk",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "sm_scale": {
+      "shape": null,
+      "dtype": "float32",
+      "description": "Model default 1/sqrt(64)."
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "B",
+        "Nq",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "B >= 1",
+    "Nq >= 1",
+    "Nk >= 1"
+  ],
+  "reference": "import math\nimport torch\n\n@torch.no_grad()\ndef run(Q, K, V, sm_scale):\n    \"\"\"Noncausal attention without a mask.\"\"\"\n    scores = torch.einsum(\"bqhd,bkhd->bhqk\", Q.float(), K.float())\n    scores = scores * float(sm_scale)\n    probabilities = torch.softmax(scores, dim=-1)\n    return torch.einsum(\"bhqk,bkhd->bqhd\", probabilities, V.float()).to(torch.bfloat16)\n"
+}
diff --git a/definitions/gemm/dense_gemm_bf16_nt.json b/definitions/gemm/dense_gemm_bf16_nt_n12288_k4096.json
similarity index 66%
rename from definitions/gemm/dense_gemm_bf16_nt.json
rename to definitions/gemm/dense_gemm_bf16_nt_n12288_k4096.json
index abf338a3cc6f197264218552b19be32ab024ae83..b5dcbe03b82859dbbe32bad295ea562aab061e41 100644
--- a/definitions/gemm/dense_gemm_bf16_nt.json
+++ b/definitions/gemm/dense_gemm_bf16_nt_n12288_k4096.json
@@ -1,7 +1,7 @@
 {
-  "name": "dense_gemm_bf16_nt",
+  "name": "dense_gemm_bf16_nt_n12288_k4096",
   "op_type": "gemm",
-  "description": "Independent A @ W.T with FP32 accumulation and BF16 output; the Qwen Image 2.1 Turbo transformer's (N, K) = (4096, 4096) Q/K/V, output and context projections, (12288, 4096) feed-forward gate/up and packed Q/K/V, and (4096, 12288) feed-forward down projections. Contiguous row-major inputs. No bias, activation, residual or quantization.",
+  "description": "Independent A @ W.T with FP32 accumulation and BF16 output; feed-forward gate/up projections. Contiguous row-major inputs. No bias, activation, residual or quantization.",
   "tags": [
     "model:qwen-image-2.1-turbo",
     "status:draft",
@@ -13,12 +13,12 @@
       "description": "Runtime number of token rows."
     },
     "N": {
-      "type": "var",
-      "description": "Output features."
+      "type": "const",
+      "value": 12288
     },
     "K": {
-      "type": "var",
-      "description": "Reduction dimension."
+      "type": "const",
+      "value": 4096
     }
   },
   "inputs": {
diff --git a/definitions/gemm/dense_gemm_bf16_nt_n4096_k12288.json b/definitions/gemm/dense_gemm_bf16_nt_n4096_k12288.json
new file mode 100644
index 0000000000000000000000000000000000000000..dda4e9d88a93f832403fbef619566ecd37649d51
--- /dev/null
+++ b/definitions/gemm/dense_gemm_bf16_nt_n4096_k12288.json
@@ -0,0 +1,53 @@
+{
+  "name": "dense_gemm_bf16_nt_n4096_k12288",
+  "op_type": "gemm",
+  "description": "Independent A @ W.T with FP32 accumulation and BF16 output; feed-forward down projection. Contiguous row-major inputs. No bias, activation, residual or quantization.",
+  "tags": [
+    "model:qwen-image-2.1-turbo",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "M": {
+      "type": "var",
+      "description": "Runtime number of token rows."
+    },
+    "N": {
+      "type": "const",
+      "value": 4096
+    },
+    "K": {
+      "type": "const",
+      "value": 12288
+    }
+  },
+  "inputs": {
+    "A": {
+      "shape": [
+        "M",
+        "K"
+      ],
+      "dtype": "bfloat16"
+    },
+    "W": {
+      "shape": [
+        "N",
+        "K"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "C": {
+      "shape": [
+        "M",
+        "N"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "M >= 1"
+  ],
+  "reference": "import math\nimport torch\n\n@torch.no_grad()\ndef run(A, W):\n    \"\"\"BF16 inputs, FP32 accumulation, one BF16 output rounding.\"\"\"\n    return (A.float() @ W.float().T).to(torch.bfloat16)\n"
+}
diff --git a/definitions/gemm/dense_gemm_bf16_nt_n4096_k4096.json b/definitions/gemm/dense_gemm_bf16_nt_n4096_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ffb01ebe4b7a6548b9d5ee26de0fa2356986789
--- /dev/null
+++ b/definitions/gemm/dense_gemm_bf16_nt_n4096_k4096.json
@@ -0,0 +1,53 @@
+{
+  "name": "dense_gemm_bf16_nt_n4096_k4096",
+  "op_type": "gemm",
+  "description": "Independent A @ W.T with FP32 accumulation and BF16 output; Q/K/V/output and context projections. Contiguous row-major inputs. No bias, activation, residual or quantization.",
+  "tags": [
+    "model:qwen-image-2.1-turbo",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "M": {
+      "type": "var",
+      "description": "Runtime number of token rows."
+    },
+    "N": {
+      "type": "const",
+      "value": 4096
+    },
+    "K": {
+      "type": "const",
+      "value": 4096
+    }
+  },
+  "inputs": {
+    "A": {
+      "shape": [
+        "M",
+        "K"
+      ],
+      "dtype": "bfloat16"
+    },
+    "W": {
+      "shape": [
+        "N",
+        "K"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "C": {
+      "shape": [
+        "M",
+        "N"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "M >= 1"
+  ],
+  "reference": "import math\nimport torch\n\n@torch.no_grad()\ndef run(A, W):\n    \"\"\"BF16 inputs, FP32 accumulation, one BF16 output rounding.\"\"\"\n    return (A.float() @ W.float().T).to(torch.bfloat16)\n"
+}
diff --git a/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.json b/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
similarity index 84%
rename from definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.json
rename to definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
index e36c59d73d5d0d4c33ac2afad7cbfaaccc668790..772116e7413248013e4b897d7eab05bdc44082c8 100644
--- a/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.json
+++ b/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
@@ -1,7 +1,7 @@
 {
-  "name": "grouped_gemm_fp8_fp4_e384_g32",
+  "name": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32",
   "op_type": "grouped_gemm",
-  "description": "Grouped E4M3 activation x packed E2M1 weight GEMM: the concatenated gate/up projection (N=4608, K=5120) and the down projection (N=5120, K=2304). Scale groups cover 32 consecutive K values. Logical unpacked scale layout, no backend swizzle. Input rows are already routed/permuted; no activation, routing weights or scatter.",
+  "description": "Grouped E4M3 activation x packed E2M1 weight GEMM, concatenated gate/up projection. Scale groups cover 32 consecutive K values. Logical unpacked scale layout, no backend swizzle. Input rows are already routed/permuted; no activation, routing weights or scatter.",
   "tags": [
     "model:deepseek-v4.1-flash",
     "status:draft",
@@ -17,20 +17,20 @@
       "value": 384
     },
     "N": {
-      "type": "var",
-      "description": "Output features."
+      "type": "const",
+      "value": 4608
     },
     "K": {
-      "type": "var",
-      "description": "Reduction dimension."
+      "type": "const",
+      "value": 5120
     },
     "K_packed": {
-      "type": "var",
-      "description": "K / 2 bytes of packed E2M1 weights."
+      "type": "const",
+      "value": 2560
     },
     "K_groups": {
-      "type": "var",
-      "description": "K / 32 scale groups."
+      "type": "const",
+      "value": 160
     },
     "group_size": {
       "type": "const",
diff --git a/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json b/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
new file mode 100644
index 0000000000000000000000000000000000000000..37e7deef0afa0b22c9692b21dd5177d90e521cab
--- /dev/null
+++ b/definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
@@ -0,0 +1,97 @@
+{
+  "name": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32",
+  "op_type": "grouped_gemm",
+  "description": "Grouped E4M3 activation x packed E2M1 weight GEMM, down projection. Scale groups cover 32 consecutive K values. Logical unpacked scale layout, no backend swizzle. Input rows are already routed/permuted; no activation, routing weights or scatter.",
+  "tags": [
+    "model:deepseek-v4.1-flash",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "M": {
+      "type": "var",
+      "description": "Runtime number of token rows."
+    },
+    "E": {
+      "type": "const",
+      "value": 384
+    },
+    "N": {
+      "type": "const",
+      "value": 5120
+    },
+    "K": {
+      "type": "const",
+      "value": 2304
+    },
+    "K_packed": {
+      "type": "const",
+      "value": 1152
+    },
+    "K_groups": {
+      "type": "const",
+      "value": 72
+    },
+    "group_size": {
+      "type": "const",
+      "value": 32
+    }
+  },
+  "inputs": {
+    "Aq": {
+      "shape": [
+        "M",
+        "K"
+      ],
+      "dtype": "float8_e4m3fn"
+    },
+    "Wq_packed": {
+      "shape": [
+        "E",
+        "N",
+        "K_packed"
+      ],
+      "dtype": "int8",
+      "description": "Raw FP4 byte carrier; even K is the low nibble."
+    },
+    "scale_A": {
+      "shape": [
+        "M",
+        "K_groups"
+      ],
+      "dtype": "int8",
+      "description": "Raw E8M0 exponent byte carrier, not signed numeric scales."
+    },
+    "scale_W": {
+      "shape": [
+        "E",
+        "N",
+        "K_groups"
+      ],
+      "dtype": "int8",
+      "description": "Raw E8M0 exponent byte carrier, not signed numeric scales."
+    },
+    "expert_id": {
+      "shape": [
+        "M"
+      ],
+      "dtype": "int32",
+      "description": "Nondecreasing expert IDs in [0,384); empty experts are allowed."
+    }
+  },
+  "outputs": {
+    "Y": {
+      "shape": [
+        "M",
+        "N"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "M >= 1",
+    "K == 2 * K_packed",
+    "K == group_size * K_groups"
+  ],
+  "reference": "import math\nimport torch\n\ndef unpack_e2m1(packed):\n    \"\"\"Decode low-nibble-first FP4 bytes without interpreting int8 numerically.\"\"\"\n    raw = packed.to(torch.int16) & 255\n    codes = torch.stack((raw & 15, raw >> 4), dim=-1).flatten(-2)\n    table = torch.tensor(\n        [0.0, 0.5, 1.0, 1.5, 2.0, 3.0, 4.0, 6.0],\n        dtype=torch.float32, device=packed.device,\n    )\n    return table[(codes & 7).long()] * torch.where((codes & 8) == 0, 1.0, -1.0)\n\n\ndef decode_e8m0(encoded):\n    \"\"\"E8M0 raw exponent bytes: 0..254 mean 2**(byte-127); 255 is invalid.\"\"\"\n    raw = encoded.to(torch.int32) & 255\n    assert bool((raw < 255).all()), \"E8M0 NaN scale is outside the contract\"\n    return torch.ldexp(torch.ones_like(raw, dtype=torch.float32), raw - 127)\n\n\n@torch.no_grad()\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    \"\"\"Rows are sorted by expert; routing and quantization are outside this op.\"\"\"\n    m, k = Aq.shape\n    experts, n, packed_k = Wq_packed.shape\n    assert k == 2 * packed_k and k % 32 == 0\n    assert bool(((expert_id >= 0) & (expert_id < experts)).all())\n    assert bool((expert_id[1:] >= expert_id[:-1]).all())\n    a = Aq.float() * decode_e8m0(scale_A).repeat_interleave(32, dim=-1)\n    result = torch.empty((m, n), dtype=torch.bfloat16, device=Aq.device)\n    for expert in torch.unique(expert_id).tolist():\n        rows = torch.nonzero(expert_id == expert, as_tuple=True)[0]\n        w = unpack_e2m1(Wq_packed[expert])\n        w = w * decode_e8m0(scale_W[expert]).repeat_interleave(32, dim=-1)\n        result[rows] = (a[rows] @ w.T).to(torch.bfloat16)\n    return result\n"
+}
diff --git a/definitions/neighborhood_attention/neighborhood_attention_3d_d64.json b/definitions/neighborhood_attention/neighborhood_attention_3d_d64.json
deleted file mode 100644
index 2aed61c097a5632476eed926bfb397937811af9c..0000000000000000000000000000000000000000
--- a/definitions/neighborhood_attention/neighborhood_attention_3d_d64.json
+++ /dev/null
@@ -1,106 +0,0 @@
-{
-  "name": "neighborhood_attention_3d_d64",
-  "op_type": "neighborhood_attention",
-  "description": "3D neighborhood attention core for the LTX-2.5 diffusion decoder. The head count and window vary by stage: (H, Kt x Ky x Kx) = (32, 3x7x7) in stage 1, (16, 3x7x7) in stage 2, (8, 3x5x5) in stages 3/4 and (4, 11x11x11) in stage 5. Q/K are normalized and rotary-embedded; Q already includes 1/sqrt(64), so no additional attention scaling occurs. Window starts clamp(p-floor(k/2),0,L-k), keeping the full window by shifting inward at boundaries. Noncausal in time. Grid dimensions must be at least the window dimensions. No auxiliary keyframe stream or enclosing block fusion.",
-  "tags": [
-    "model:ltx-2.5",
-    "status:draft",
-    "parallelism:unsharded"
-  ],
-  "axes": {
-    "B": {
-      "type": "var",
-      "description": "Runtime batch size."
-    },
-    "T": {
-      "type": "var",
-      "description": "Runtime frame count."
-    },
-    "Y": {
-      "type": "var",
-      "description": "Runtime grid height."
-    },
-    "X": {
-      "type": "var",
-      "description": "Runtime grid width."
-    },
-    "H": {
-      "type": "var",
-      "description": "Attention heads of the decoder stage."
-    },
-    "D": {
-      "type": "const",
-      "value": 64
-    }
-  },
-  "inputs": {
-    "Q": {
-      "shape": [
-        "B",
-        "T",
-        "Y",
-        "X",
-        "H",
-        "D"
-      ],
-      "dtype": "bfloat16"
-    },
-    "K": {
-      "shape": [
-        "B",
-        "T",
-        "Y",
-        "X",
-        "H",
-        "D"
-      ],
-      "dtype": "bfloat16"
-    },
-    "V": {
-      "shape": [
-        "B",
-        "T",
-        "Y",
-        "X",
-        "H",
-        "D"
-      ],
-      "dtype": "bfloat16"
-    },
-    "Kt": {
-      "shape": null,
-      "dtype": "int32",
-      "description": "Window extent (t)."
-    },
-    "Ky": {
-      "shape": null,
-      "dtype": "int32",
-      "description": "Window extent (y)."
-    },
-    "Kx": {
-      "shape": null,
-      "dtype": "int32",
-      "description": "Window extent (x)."
-    }
-  },
-  "outputs": {
-    "O": {
-      "shape": [
-        "B",
-        "T",
-        "Y",
-        "X",
-        "H",
-        "D"
-      ],
-      "dtype": "bfloat16"
-    }
-  },
-  "constraints": [
-    "B >= 1",
-    "T >= Kt",
-    "Y >= Ky",
-    "X >= Kx"
-  ],
-  "reference": "import math\nimport torch\n\n@torch.no_grad()\ndef run(Q, K, V, Kt, Ky, Kx):\n    \"\"\"Q is pre-scaled; shift complete fixed-size windows inward at borders.\"\"\"\n    kt, ky, kx = int(Kt), int(Ky), int(Kx)\n    bsz, frames, height, width, heads, dim = Q.shape\n    assert frames >= kt and height >= ky and width >= kx\n    qflat, kflat, vflat = (x.flatten(1, 3) for x in (Q, K, V))\n    out = torch.empty_like(qflat)\n    offsets = (\n        torch.arange(kt, device=Q.device)[:, None, None] * height * width\n        + torch.arange(ky, device=Q.device)[None, :, None] * width\n        + torch.arange(kx, device=Q.device)[None, None, :]\n    ).flatten()\n    # Bound gathered-neighborhood memory independently of the full video volume.\n    for first in range(0, qflat.shape[1], 32):\n        rows = torch.arange(first, min(first + 32, qflat.shape[1]), device=Q.device)\n        t, y, x = rows // (height * width), (rows // width) % height, rows % width\n        start = ((t - kt // 2).clamp(0, frames - kt) * height * width\n                 + (y - ky // 2).clamp(0, height - ky) * width\n                 + (x - kx // 2).clamp(0, width - kx))\n        neighbors = start[:, None] + offsets[None, :]\n        keys, values = kflat[:, neighbors].float(), vflat[:, neighbors].float()\n        scores = torch.einsum(\"bqhd,bqkhd->bqhk\", qflat[:, rows].float(), keys)\n        out[:, rows] = torch.einsum(\n            \"bqhk,bqkhd->bqhd\", scores.softmax(-1), values\n        ).to(torch.bfloat16)\n    return out.reshape_as(Q)\n"
-}
diff --git a/definitions/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json b/definitions/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8f83ee869040de251842b61c269e219e883a8fc
--- /dev/null
+++ b/definitions/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
@@ -0,0 +1,103 @@
+{
+  "name": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7",
+  "op_type": "neighborhood_attention",
+  "description": "3D neighborhood attention core for LTX-2.5 diffusion decoder stages 2. Q/K are normalized and rotary-embedded; Q already includes 1/sqrt(64), so no additional attention scaling occurs. Window starts clamp(p-floor(k/2),0,L-k), keeping the full window by shifting inward at boundaries. Noncausal in time. Grid dimensions must be at least the window dimensions. No auxiliary keyframe stream or enclosing block fusion.",
+  "tags": [
+    "model:ltx-2.5",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "B": {
+      "type": "var",
+      "description": "Runtime batch size."
+    },
+    "T": {
+      "type": "var",
+      "description": "Runtime frame count."
+    },
+    "Y": {
+      "type": "var",
+      "description": "Runtime grid height."
+    },
+    "X": {
+      "type": "var",
+      "description": "Runtime grid width."
+    },
+    "H": {
+      "type": "const",
+      "value": 16
+    },
+    "D": {
+      "type": "const",
+      "value": 64
+    },
+    "Kt": {
+      "type": "const",
+      "value": 3
+    },
+    "Ky": {
+      "type": "const",
+      "value": 7
+    },
+    "Kx": {
+      "type": "const",
+      "value": 7
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "K": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "V": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "B >= 1",
+    "T >= Kt",
+    "Y >= Ky",
+    "X >= Kx"
+  ],
+  "reference": "import math\nimport torch\n\nWINDOW = (3, 7, 7)\n\n@torch.no_grad()\ndef run(Q, K, V):\n    \"\"\"Q is pre-scaled; shift complete fixed-size windows inward at borders.\"\"\"\n    kt, ky, kx = WINDOW\n    bsz, frames, height, width, heads, dim = Q.shape\n    assert frames >= kt and height >= ky and width >= kx\n    qflat, kflat, vflat = (x.flatten(1, 3) for x in (Q, K, V))\n    out = torch.empty_like(qflat)\n    offsets = (\n        torch.arange(kt, device=Q.device)[:, None, None] * height * width\n        + torch.arange(ky, device=Q.device)[None, :, None] * width\n        + torch.arange(kx, device=Q.device)[None, None, :]\n    ).flatten()\n    # Bound gathered-neighborhood memory independently of the full video volume.\n    for first in range(0, qflat.shape[1], 32):\n        rows = torch.arange(first, min(first + 32, qflat.shape[1]), device=Q.device)\n        t, y, x = rows // (height * width), (rows // width) % height, rows % width\n        start = ((t - kt // 2).clamp(0, frames - kt) * height * width\n                 + (y - ky // 2).clamp(0, height - ky) * width\n                 + (x - kx // 2).clamp(0, width - kx))\n        neighbors = start[:, None] + offsets[None, :]\n        keys, values = kflat[:, neighbors].float(), vflat[:, neighbors].float()\n        scores = torch.einsum(\"bqhd,bqkhd->bqhk\", qflat[:, rows].float(), keys)\n        out[:, rows] = torch.einsum(\n            \"bqhk,bqkhd->bqhd\", scores.softmax(-1), values\n        ).to(torch.bfloat16)\n    return out.reshape_as(Q)\n"
+}
diff --git a/definitions/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json b/definitions/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a9330586857b96517b11f8efca7307e6c5d148d
--- /dev/null
+++ b/definitions/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
@@ -0,0 +1,103 @@
+{
+  "name": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7",
+  "op_type": "neighborhood_attention",
+  "description": "3D neighborhood attention core for LTX-2.5 diffusion decoder stages 1. Q/K are normalized and rotary-embedded; Q already includes 1/sqrt(64), so no additional attention scaling occurs. Window starts clamp(p-floor(k/2),0,L-k), keeping the full window by shifting inward at boundaries. Noncausal in time. Grid dimensions must be at least the window dimensions. No auxiliary keyframe stream or enclosing block fusion.",
+  "tags": [
+    "model:ltx-2.5",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "B": {
+      "type": "var",
+      "description": "Runtime batch size."
+    },
+    "T": {
+      "type": "var",
+      "description": "Runtime frame count."
+    },
+    "Y": {
+      "type": "var",
+      "description": "Runtime grid height."
+    },
+    "X": {
+      "type": "var",
+      "description": "Runtime grid width."
+    },
+    "H": {
+      "type": "const",
+      "value": 32
+    },
+    "D": {
+      "type": "const",
+      "value": 64
+    },
+    "Kt": {
+      "type": "const",
+      "value": 3
+    },
+    "Ky": {
+      "type": "const",
+      "value": 7
+    },
+    "Kx": {
+      "type": "const",
+      "value": 7
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "K": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "V": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "B >= 1",
+    "T >= Kt",
+    "Y >= Ky",
+    "X >= Kx"
+  ],
+  "reference": "import math\nimport torch\n\nWINDOW = (3, 7, 7)\n\n@torch.no_grad()\ndef run(Q, K, V):\n    \"\"\"Q is pre-scaled; shift complete fixed-size windows inward at borders.\"\"\"\n    kt, ky, kx = WINDOW\n    bsz, frames, height, width, heads, dim = Q.shape\n    assert frames >= kt and height >= ky and width >= kx\n    qflat, kflat, vflat = (x.flatten(1, 3) for x in (Q, K, V))\n    out = torch.empty_like(qflat)\n    offsets = (\n        torch.arange(kt, device=Q.device)[:, None, None] * height * width\n        + torch.arange(ky, device=Q.device)[None, :, None] * width\n        + torch.arange(kx, device=Q.device)[None, None, :]\n    ).flatten()\n    # Bound gathered-neighborhood memory independently of the full video volume.\n    for first in range(0, qflat.shape[1], 32):\n        rows = torch.arange(first, min(first + 32, qflat.shape[1]), device=Q.device)\n        t, y, x = rows // (height * width), (rows // width) % height, rows % width\n        start = ((t - kt // 2).clamp(0, frames - kt) * height * width\n                 + (y - ky // 2).clamp(0, height - ky) * width\n                 + (x - kx // 2).clamp(0, width - kx))\n        neighbors = start[:, None] + offsets[None, :]\n        keys, values = kflat[:, neighbors].float(), vflat[:, neighbors].float()\n        scores = torch.einsum(\"bqhd,bqkhd->bqhk\", qflat[:, rows].float(), keys)\n        out[:, rows] = torch.einsum(\n            \"bqhk,bqkhd->bqhd\", scores.softmax(-1), values\n        ).to(torch.bfloat16)\n    return out.reshape_as(Q)\n"
+}
diff --git a/definitions/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json b/definitions/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
new file mode 100644
index 0000000000000000000000000000000000000000..5716a90ff00bff0b0f2b6e52022b608278de99e8
--- /dev/null
+++ b/definitions/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
@@ -0,0 +1,103 @@
+{
+  "name": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11",
+  "op_type": "neighborhood_attention",
+  "description": "3D neighborhood attention core for LTX-2.5 diffusion decoder stages 5. Q/K are normalized and rotary-embedded; Q already includes 1/sqrt(64), so no additional attention scaling occurs. Window starts clamp(p-floor(k/2),0,L-k), keeping the full window by shifting inward at boundaries. Noncausal in time. Grid dimensions must be at least the window dimensions. No auxiliary keyframe stream or enclosing block fusion.",
+  "tags": [
+    "model:ltx-2.5",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "B": {
+      "type": "var",
+      "description": "Runtime batch size."
+    },
+    "T": {
+      "type": "var",
+      "description": "Runtime frame count."
+    },
+    "Y": {
+      "type": "var",
+      "description": "Runtime grid height."
+    },
+    "X": {
+      "type": "var",
+      "description": "Runtime grid width."
+    },
+    "H": {
+      "type": "const",
+      "value": 4
+    },
+    "D": {
+      "type": "const",
+      "value": 64
+    },
+    "Kt": {
+      "type": "const",
+      "value": 11
+    },
+    "Ky": {
+      "type": "const",
+      "value": 11
+    },
+    "Kx": {
+      "type": "const",
+      "value": 11
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "K": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "V": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "B >= 1",
+    "T >= Kt",
+    "Y >= Ky",
+    "X >= Kx"
+  ],
+  "reference": "import math\nimport torch\n\nWINDOW = (11, 11, 11)\n\n@torch.no_grad()\ndef run(Q, K, V):\n    \"\"\"Q is pre-scaled; shift complete fixed-size windows inward at borders.\"\"\"\n    kt, ky, kx = WINDOW\n    bsz, frames, height, width, heads, dim = Q.shape\n    assert frames >= kt and height >= ky and width >= kx\n    qflat, kflat, vflat = (x.flatten(1, 3) for x in (Q, K, V))\n    out = torch.empty_like(qflat)\n    offsets = (\n        torch.arange(kt, device=Q.device)[:, None, None] * height * width\n        + torch.arange(ky, device=Q.device)[None, :, None] * width\n        + torch.arange(kx, device=Q.device)[None, None, :]\n    ).flatten()\n    # Bound gathered-neighborhood memory independently of the full video volume.\n    for first in range(0, qflat.shape[1], 32):\n        rows = torch.arange(first, min(first + 32, qflat.shape[1]), device=Q.device)\n        t, y, x = rows // (height * width), (rows // width) % height, rows % width\n        start = ((t - kt // 2).clamp(0, frames - kt) * height * width\n                 + (y - ky // 2).clamp(0, height - ky) * width\n                 + (x - kx // 2).clamp(0, width - kx))\n        neighbors = start[:, None] + offsets[None, :]\n        keys, values = kflat[:, neighbors].float(), vflat[:, neighbors].float()\n        scores = torch.einsum(\"bqhd,bqkhd->bqhk\", qflat[:, rows].float(), keys)\n        out[:, rows] = torch.einsum(\n            \"bqhk,bqkhd->bqhd\", scores.softmax(-1), values\n        ).to(torch.bfloat16)\n    return out.reshape_as(Q)\n"
+}
diff --git a/definitions/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json b/definitions/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
new file mode 100644
index 0000000000000000000000000000000000000000..20bde0c98793ba08b0cdb135f736030f6dbc42ba
--- /dev/null
+++ b/definitions/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
@@ -0,0 +1,103 @@
+{
+  "name": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5",
+  "op_type": "neighborhood_attention",
+  "description": "3D neighborhood attention core for LTX-2.5 diffusion decoder stages 3/4. Q/K are normalized and rotary-embedded; Q already includes 1/sqrt(64), so no additional attention scaling occurs. Window starts clamp(p-floor(k/2),0,L-k), keeping the full window by shifting inward at boundaries. Noncausal in time. Grid dimensions must be at least the window dimensions. No auxiliary keyframe stream or enclosing block fusion.",
+  "tags": [
+    "model:ltx-2.5",
+    "status:draft",
+    "parallelism:unsharded"
+  ],
+  "axes": {
+    "B": {
+      "type": "var",
+      "description": "Runtime batch size."
+    },
+    "T": {
+      "type": "var",
+      "description": "Runtime frame count."
+    },
+    "Y": {
+      "type": "var",
+      "description": "Runtime grid height."
+    },
+    "X": {
+      "type": "var",
+      "description": "Runtime grid width."
+    },
+    "H": {
+      "type": "const",
+      "value": 8
+    },
+    "D": {
+      "type": "const",
+      "value": 64
+    },
+    "Kt": {
+      "type": "const",
+      "value": 3
+    },
+    "Ky": {
+      "type": "const",
+      "value": 5
+    },
+    "Kx": {
+      "type": "const",
+      "value": 5
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "K": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "V": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "B",
+        "T",
+        "Y",
+        "X",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "B >= 1",
+    "T >= Kt",
+    "Y >= Ky",
+    "X >= Kx"
+  ],
+  "reference": "import math\nimport torch\n\nWINDOW = (3, 5, 5)\n\n@torch.no_grad()\ndef run(Q, K, V):\n    \"\"\"Q is pre-scaled; shift complete fixed-size windows inward at borders.\"\"\"\n    kt, ky, kx = WINDOW\n    bsz, frames, height, width, heads, dim = Q.shape\n    assert frames >= kt and height >= ky and width >= kx\n    qflat, kflat, vflat = (x.flatten(1, 3) for x in (Q, K, V))\n    out = torch.empty_like(qflat)\n    offsets = (\n        torch.arange(kt, device=Q.device)[:, None, None] * height * width\n        + torch.arange(ky, device=Q.device)[None, :, None] * width\n        + torch.arange(kx, device=Q.device)[None, None, :]\n    ).flatten()\n    # Bound gathered-neighborhood memory independently of the full video volume.\n    for first in range(0, qflat.shape[1], 32):\n        rows = torch.arange(first, min(first + 32, qflat.shape[1]), device=Q.device)\n        t, y, x = rows // (height * width), (rows // width) % height, rows % width\n        start = ((t - kt // 2).clamp(0, frames - kt) * height * width\n                 + (y - ky // 2).clamp(0, height - ky) * width\n                 + (x - kx // 2).clamp(0, width - kx))\n        neighbors = start[:, None] + offsets[None, :]\n        keys, values = kflat[:, neighbors].float(), vflat[:, neighbors].float()\n        scores = torch.einsum(\"bqhd,bqkhd->bqhk\", qflat[:, rows].float(), keys)\n        out[:, rows] = torch.einsum(\n            \"bqhk,bqkhd->bqhd\", scores.softmax(-1), values\n        ).to(torch.bfloat16)\n    return out.reshape_as(Q)\n"
+}
diff --git a/definitions/sparse_mla/sparse_mla_prefill_h64_d512.json b/definitions/sparse_mla/sparse_mla_prefill_h64_d512_k128.json
similarity index 96%
rename from definitions/sparse_mla/sparse_mla_prefill_h64_d512.json
rename to definitions/sparse_mla/sparse_mla_prefill_h64_d512_k128.json
index 9ce351336f88b10e7714bd21ec99cbf10cc8c1a7..f4ac089f1f189fe94377f70c5d65d558f5ddc7ca 100644
--- a/definitions/sparse_mla/sparse_mla_prefill_h64_d512.json
+++ b/definitions/sparse_mla/sparse_mla_prefill_h64_d512_k128.json
@@ -1,5 +1,5 @@
 {
-  "name": "sparse_mla_prefill_h64_d512",
+  "name": "sparse_mla_prefill_h64_d512_k128",
   "op_type": "sparse_mla",
   "description": "Sparse Multi-head Latent Attention for prefill, with BF16 Q and a BF16 latent workspace. K and V are the same latent vector. Each query has its own sparse index list and length; visibility is fully specified by these inputs, with no additional causal mask. One softmax includes all valid selected slots and a per-head zero-value sink. Queries with no selected keys return zero. Cache dequantization, index selection, Q normalization/rotary embedding and output projection are outside this operator.",
   "tags": [
@@ -30,8 +30,8 @@
       "value": 1
     },
     "TopK": {
-      "type": "var",
-      "description": "Selected slots per query: 128 or 640."
+      "type": "const",
+      "value": 128
     }
   },
   "inputs": {
diff --git a/definitions/sparse_mla/sparse_mla_prefill_h64_d512_k640.json b/definitions/sparse_mla/sparse_mla_prefill_h64_d512_k640.json
new file mode 100644
index 0000000000000000000000000000000000000000..4614d82bd0554ce966ceb30d7b8cec0585c3f538
--- /dev/null
+++ b/definitions/sparse_mla/sparse_mla_prefill_h64_d512_k640.json
@@ -0,0 +1,98 @@
+{
+  "name": "sparse_mla_prefill_h64_d512_k640",
+  "op_type": "sparse_mla",
+  "description": "Sparse Multi-head Latent Attention for prefill, with BF16 Q and a BF16 latent workspace. K and V are the same latent vector. Each query has its own sparse index list and length; visibility is fully specified by these inputs, with no additional causal mask. One softmax includes all valid selected slots and a per-head zero-value sink. Queries with no selected keys return zero. Cache dequantization, index selection, Q normalization/rotary embedding and output projection are outside this operator.",
+  "tags": [
+    "model:deepseek-v4.1-flash",
+    "status:draft",
+    "parallelism:unsharded",
+    "stage:prefill"
+  ],
+  "axes": {
+    "M": {
+      "type": "var",
+      "description": "Runtime number of token rows."
+    },
+    "Nkv": {
+      "type": "var",
+      "description": "Number of latent vectors in the engine workspace."
+    },
+    "H": {
+      "type": "const",
+      "value": 64
+    },
+    "D": {
+      "type": "const",
+      "value": 512
+    },
+    "KVH": {
+      "type": "const",
+      "value": 1
+    },
+    "TopK": {
+      "type": "const",
+      "value": 640
+    }
+  },
+  "inputs": {
+    "Q": {
+      "shape": [
+        "M",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "KV": {
+      "shape": [
+        "Nkv",
+        "KVH",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    },
+    "indices": {
+      "shape": [
+        "M",
+        "KVH",
+        "TopK"
+      ],
+      "dtype": "int32",
+      "description": "Physical workspace indices; -1 or indices >= Nkv are ignored."
+    },
+    "topk_length": {
+      "shape": [
+        "M"
+      ],
+      "dtype": "int32",
+      "description": "Only slots before this per-query length are considered."
+    },
+    "attn_sink": {
+      "shape": [
+        "H"
+      ],
+      "dtype": "float32",
+      "description": "Per-head zero-value sink logit; -inf disables, +inf forces zero output."
+    },
+    "sm_scale": {
+      "shape": null,
+      "dtype": "float32",
+      "description": "Model default 1/sqrt(512)."
+    }
+  },
+  "outputs": {
+    "O": {
+      "shape": [
+        "M",
+        "H",
+        "D"
+      ],
+      "dtype": "bfloat16"
+    }
+  },
+  "constraints": [
+    "M >= 1",
+    "Nkv >= 1"
+  ],
+  "reference": "import math\nimport torch\n\n@torch.no_grad()\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    \"\"\"Per-query sparse BF16 latent attention, including the zero-value sink.\"\"\"\n    result = torch.empty_like(Q)\n    slots = torch.arange(indices.shape[-1], device=Q.device)\n    sink = attn_sink.float()[None, :, None]\n    for start in range(0, Q.shape[0], 16):\n        end = min(start + 16, Q.shape[0])\n        ids = indices[start:end, 0].long()\n        valid = ((ids >= 0) & (ids < KV.shape[0]) &\n                 (slots[None] < topk_length[start:end, None]))\n        values = KV[ids.clamp(0, KV.shape[0] - 1), 0].float()\n        values = values.masked_fill(~valid[..., None], 0)\n        logits = torch.bmm(Q[start:end].float(), values.transpose(1, 2)) * float(sm_scale)\n        logits.masked_fill_(~valid[:, None], -torch.inf)\n        # Empty rows return zero even when the sink is disabled with -inf.\n        empty = ~valid.any(-1)\n        row_sink = sink.expand(end - start, -1, -1)\n        row_sink = torch.where(empty[:, None, None], 0.0, row_sink)\n        probabilities = torch.softmax(torch.cat((logits, row_sink), -1), -1)[..., :-1]\n        probabilities = torch.where(torch.isposinf(sink), 0.0, probabilities)\n        result[start:end] = torch.bmm(probabilities, values).to(Q.dtype)\n    return result\n"
+}
diff --git a/docs/shape-distributions/README.md b/docs/shape-distributions/README.md
index 0e640434a40b4ebf648bc385455324d3bf789dea..3bfd477a20ea5f2ee48ac3792b21e6753fa78d16 100644
--- a/docs/shape-distributions/README.md
+++ b/docs/shape-distributions/README.md
@@ -1,6 +1,6 @@
 # Kernel shape distributions
 
-**13 kernels, 1104 captured workloads.** Each histogram counts the workload records in one kernel definition's JSONL file. Every bar is one exact combination of its runtime dimensions; identical shapes are combined. The labels retain all runtime dimensions, including dimensions that happen to have one observed value. Model constants appear above each plot.
+**21 kernels, 1104 captured workloads.** Each histogram counts the workload records in one kernel definition's JSONL file. Every bar is one exact combination of its runtime dimensions; identical shapes are combined. The labels retain all runtime dimensions, including dimensions that happen to have one observed value. Model constants appear above each plot.
 
 Read bar height (or length in horizontal plots) as the number of captured workloads with that shape. Bins are equally spaced categories, sorted by the dimension values, not continuous numeric intervals. A bar of 1 means one published example; a bar of 6 means six. These are capture counts, not request probabilities, runtime invocation frequencies, or AgentX traffic weights.
 
@@ -18,17 +18,25 @@ Only definition and workload metadata are needed; tensor snapshots and a GPU are
 |---|---:|---:|
 | [prefix_cached_block_causal_h32_d128](#prefix_cached_block_causal_h32_d128) | 28 | 4 |
 | [prefix_cached_block_causal_h32_d128_prefix0](#prefix_cached_block_causal_h32_d128_prefix0) | 12 | 4 |
-| [noncausal_dense_attention_h32](#noncausal_dense_attention_h32) | 162 | 48 |
+| [noncausal_dense_attention_h32_d128](#noncausal_dense_attention_h32_d128) | 54 | 18 |
+| [noncausal_dense_attention_h32_d64](#noncausal_dense_attention_h32_d64) | 108 | 30 |
 | [gdn_decode_qk4_v8_d128_k_last](#gdn_decode_qk4_v8_d128_k_last) | 54 | 7 |
 | [gdn_prefill_qk4_v8_d128_k_last](#gdn_prefill_qk4_v8_d128_k_last) | 100 | 77 |
-| [dense_gemm_bf16_nt](#dense_gemm_bf16_nt) | 93 | 21 |
-| [grouped_gemm_fp8_fp4_e384_g32](#grouped_gemm_fp8_fp4_e384_g32) | 104 | 22 |
+| [dense_gemm_bf16_nt_n12288_k4096](#dense_gemm_bf16_nt_n12288_k4096) | 31 | 7 |
+| [dense_gemm_bf16_nt_n4096_k12288](#dense_gemm_bf16_nt_n4096_k12288) | 31 | 7 |
+| [dense_gemm_bf16_nt_n4096_k4096](#dense_gemm_bf16_nt_n4096_k4096) | 31 | 7 |
+| [grouped_gemm_fp8_fp4_e384_n4608_k5120_g32](#grouped_gemm_fp8_fp4_e384_n4608_k5120_g32) | 52 | 11 |
+| [grouped_gemm_fp8_fp4_e384_n5120_k2304_g32](#grouped_gemm_fp8_fp4_e384_n5120_k2304_g32) | 52 | 11 |
 | [mixed_mla_decode_deepseek41_h64_d512_k8128_k4512](#mixed_mla_decode_deepseek41_h64_d512_k8128_k4512) | 42 | 30 |
 | [moe_router_gemm_topk_deepseek41_d5120_e384_topk6](#moe_router_gemm_topk_deepseek41_d5120_e384_topk6) | 50 | 11 |
-| [neighborhood_attention_3d_d64](#neighborhood_attention_3d_d64) | 45 | 45 |
+| [neighborhood_attention_3d_h16_d64_kt3_ky7_kx7](#neighborhood_attention_3d_h16_d64_kt3_ky7_kx7) | 9 | 9 |
+| [neighborhood_attention_3d_h32_d64_kt3_ky7_kx7](#neighborhood_attention_3d_h32_d64_kt3_ky7_kx7) | 9 | 9 |
+| [neighborhood_attention_3d_h4_d64_kt11_ky11_kx11](#neighborhood_attention_3d_h4_d64_kt11_ky11_kx11) | 9 | 9 |
+| [neighborhood_attention_3d_h8_d64_kt3_ky5_kx5](#neighborhood_attention_3d_h8_d64_kt3_ky5_kx5) | 18 | 18 |
 | [qsa_index_topk_decode_qwen38_h4_d128_topk512_c4](#qsa_index_topk_decode_qwen38_h4_d128_topk512_c4) | 192 | 77 |
 | [qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4](#qsa_sparse_gqa_decode_qwen38_h24_kv2_d256_topk512_c4) | 192 | 64 |
-| [sparse_mla_prefill_h64_d512](#sparse_mla_prefill_h64_d512) | 30 | 29 |
+| [sparse_mla_prefill_h64_d512_k128](#sparse_mla_prefill_h64_d512_k128) | 10 | 9 |
+| [sparse_mla_prefill_h64_d512_k640](#sparse_mla_prefill_h64_d512_k640) | 20 | 20 |
 
 ## prefix_cached_block_causal_h32_d128
 
@@ -46,13 +54,21 @@ Only definition and workload metadata are needed; tensor snapshots and a GPU are
 
 **Runtime dimensions:** `B`: Runtime batch size. `Nq`: Runtime current query/token count. `Nall`: Total prefix and current length.
 
-## noncausal_dense_attention_h32
+## noncausal_dense_attention_h32_d128
 
-[Definition](../../definitions/dense_attention/noncausal_dense_attention_h32.json) · [Workloads](../../workloads/dense_attention/noncausal_dense_attention_h32.jsonl) · [SVG](noncausal_dense_attention_h32.svg)
+[Definition](../../definitions/dense_attention/noncausal_dense_attention_h32_d128.json) · [Workloads](../../workloads/dense_attention/noncausal_dense_attention_h32_d128.jsonl) · [SVG](noncausal_dense_attention_h32_d128.svg)
 
-![Shape distribution for noncausal_dense_attention_h32: 162 workloads across 48 shapes](noncausal_dense_attention_h32.svg)
+![Shape distribution for noncausal_dense_attention_h32_d128: 54 workloads across 18 shapes](noncausal_dense_attention_h32_d128.svg)
 
-**Runtime dimensions:** `B`: Runtime batch size. `Nq`: Runtime query length. `Nk`: Runtime key/value length. `D`: Head dimension: 64 for audio, 128 for video.
+**Runtime dimensions:** `B`: Runtime batch size. `Nq`: Runtime query length. `Nk`: Runtime key/value length.
+
+## noncausal_dense_attention_h32_d64
+
+[Definition](../../definitions/dense_attention/noncausal_dense_attention_h32_d64.json) · [Workloads](../../workloads/dense_attention/noncausal_dense_attention_h32_d64.jsonl) · [SVG](noncausal_dense_attention_h32_d64.svg)
+
+![Shape distribution for noncausal_dense_attention_h32_d64: 108 workloads across 30 shapes](noncausal_dense_attention_h32_d64.svg)
+
+**Runtime dimensions:** `B`: Runtime batch size. `Nq`: Runtime query length. `Nk`: Runtime key/value length.
 
 ## gdn_decode_qk4_v8_d128_k_last
 
@@ -70,21 +86,45 @@ Only definition and workload metadata are needed; tensor snapshots and a GPU are
 
 **Runtime dimensions:** `total_seq_len`: total_seq_len `num_seqs`: num_seqs `len_cu_seqlens`: Length of cu_seqlens array (num_seqs + 1).
 
-## dense_gemm_bf16_nt
+## dense_gemm_bf16_nt_n12288_k4096
+
+[Definition](../../definitions/gemm/dense_gemm_bf16_nt_n12288_k4096.json) · [Workloads](../../workloads/gemm/dense_gemm_bf16_nt_n12288_k4096.jsonl) · [SVG](dense_gemm_bf16_nt_n12288_k4096.svg)
 
-[Definition](../../definitions/gemm/dense_gemm_bf16_nt.json) · [Workloads](../../workloads/gemm/dense_gemm_bf16_nt.jsonl) · [SVG](dense_gemm_bf16_nt.svg)
+![Shape distribution for dense_gemm_bf16_nt_n12288_k4096: 31 workloads across 7 shapes](dense_gemm_bf16_nt_n12288_k4096.svg)
 
-![Shape distribution for dense_gemm_bf16_nt: 93 workloads across 21 shapes](dense_gemm_bf16_nt.svg)
+**Runtime dimensions:** `M`: Runtime number of token rows.
+
+## dense_gemm_bf16_nt_n4096_k12288
+
+[Definition](../../definitions/gemm/dense_gemm_bf16_nt_n4096_k12288.json) · [Workloads](../../workloads/gemm/dense_gemm_bf16_nt_n4096_k12288.jsonl) · [SVG](dense_gemm_bf16_nt_n4096_k12288.svg)
 
-**Runtime dimensions:** `M`: Runtime number of token rows. `N`: Output features. `K`: Reduction dimension.
+![Shape distribution for dense_gemm_bf16_nt_n4096_k12288: 31 workloads across 7 shapes](dense_gemm_bf16_nt_n4096_k12288.svg)
+
+**Runtime dimensions:** `M`: Runtime number of token rows.
 
-## grouped_gemm_fp8_fp4_e384_g32
+## dense_gemm_bf16_nt_n4096_k4096
+
+[Definition](../../definitions/gemm/dense_gemm_bf16_nt_n4096_k4096.json) · [Workloads](../../workloads/gemm/dense_gemm_bf16_nt_n4096_k4096.jsonl) · [SVG](dense_gemm_bf16_nt_n4096_k4096.svg)
+
+![Shape distribution for dense_gemm_bf16_nt_n4096_k4096: 31 workloads across 7 shapes](dense_gemm_bf16_nt_n4096_k4096.svg)
+
+**Runtime dimensions:** `M`: Runtime number of token rows.
+
+## grouped_gemm_fp8_fp4_e384_n4608_k5120_g32
+
+[Definition](../../definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json) · [Workloads](../../workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.jsonl) · [SVG](grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.svg)
+
+![Shape distribution for grouped_gemm_fp8_fp4_e384_n4608_k5120_g32: 52 workloads across 11 shapes](grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.svg)
+
+**Runtime dimensions:** `M`: Runtime number of token rows.
 
-[Definition](../../definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.json) · [Workloads](../../workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.jsonl) · [SVG](grouped_gemm_fp8_fp4_e384_g32.svg)
+## grouped_gemm_fp8_fp4_e384_n5120_k2304_g32
 
-![Shape distribution for grouped_gemm_fp8_fp4_e384_g32: 104 workloads across 22 shapes](grouped_gemm_fp8_fp4_e384_g32.svg)
+[Definition](../../definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json) · [Workloads](../../workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.jsonl) · [SVG](grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.svg)
 
-**Runtime dimensions:** `M`: Runtime number of token rows. `N`: Output features. `K`: Reduction dimension. `K_packed`: K / 2 bytes of packed E2M1 weights. `K_groups`: K / 32 scale groups.
+![Shape distribution for grouped_gemm_fp8_fp4_e384_n5120_k2304_g32: 52 workloads across 11 shapes](grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.svg)
+
+**Runtime dimensions:** `M`: Runtime number of token rows.
 
 ## mixed_mla_decode_deepseek41_h64_d512_k8128_k4512
 
@@ -102,13 +142,37 @@ Only definition and workload metadata are needed; tensor snapshots and a GPU are
 
 **Runtime dimensions:** `M`: Runtime number of token rows.
 
-## neighborhood_attention_3d_d64
+## neighborhood_attention_3d_h16_d64_kt3_ky7_kx7
+
+[Definition](../../definitions/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json) · [Workloads](../../workloads/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.jsonl) · [SVG](neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.svg)
+
+![Shape distribution for neighborhood_attention_3d_h16_d64_kt3_ky7_kx7: 9 workloads across 9 shapes](neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.svg)
+
+**Runtime dimensions:** `B`: Runtime batch size. `T`: Runtime frame count. `Y`: Runtime grid height. `X`: Runtime grid width.
+
+## neighborhood_attention_3d_h32_d64_kt3_ky7_kx7
+
+[Definition](../../definitions/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json) · [Workloads](../../workloads/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.jsonl) · [SVG](neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.svg)
+
+![Shape distribution for neighborhood_attention_3d_h32_d64_kt3_ky7_kx7: 9 workloads across 9 shapes](neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.svg)
+
+**Runtime dimensions:** `B`: Runtime batch size. `T`: Runtime frame count. `Y`: Runtime grid height. `X`: Runtime grid width.
 
-[Definition](../../definitions/neighborhood_attention/neighborhood_attention_3d_d64.json) · [Workloads](../../workloads/neighborhood_attention/neighborhood_attention_3d_d64.jsonl) · [SVG](neighborhood_attention_3d_d64.svg)
+## neighborhood_attention_3d_h4_d64_kt11_ky11_kx11
 
-![Shape distribution for neighborhood_attention_3d_d64: 45 workloads across 45 shapes](neighborhood_attention_3d_d64.svg)
+[Definition](../../definitions/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json) · [Workloads](../../workloads/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.jsonl) · [SVG](neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.svg)
 
-**Runtime dimensions:** `B`: Runtime batch size. `T`: Runtime frame count. `Y`: Runtime grid height. `X`: Runtime grid width. `H`: Attention heads of the decoder stage.
+![Shape distribution for neighborhood_attention_3d_h4_d64_kt11_ky11_kx11: 9 workloads across 9 shapes](neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.svg)
+
+**Runtime dimensions:** `B`: Runtime batch size. `T`: Runtime frame count. `Y`: Runtime grid height. `X`: Runtime grid width.
+
+## neighborhood_attention_3d_h8_d64_kt3_ky5_kx5
+
+[Definition](../../definitions/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json) · [Workloads](../../workloads/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.jsonl) · [SVG](neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.svg)
+
+![Shape distribution for neighborhood_attention_3d_h8_d64_kt3_ky5_kx5: 18 workloads across 18 shapes](neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.svg)
+
+**Runtime dimensions:** `B`: Runtime batch size. `T`: Runtime frame count. `Y`: Runtime grid height. `X`: Runtime grid width.
 
 ## qsa_index_topk_decode_qwen38_h4_d128_topk512_c4
 
@@ -126,10 +190,18 @@ Only definition and workload metadata are needed; tensor snapshots and a GPU are
 
 **Runtime dimensions:** `B`: Runtime batch size. `NP`: Allocated cache page count. `P`: Backend token-cache page size. `MP`: Page-table capacity per sequence.
 
-## sparse_mla_prefill_h64_d512
+## sparse_mla_prefill_h64_d512_k128
+
+[Definition](../../definitions/sparse_mla/sparse_mla_prefill_h64_d512_k128.json) · [Workloads](../../workloads/sparse_mla/sparse_mla_prefill_h64_d512_k128.jsonl) · [SVG](sparse_mla_prefill_h64_d512_k128.svg)
+
+![Shape distribution for sparse_mla_prefill_h64_d512_k128: 10 workloads across 9 shapes](sparse_mla_prefill_h64_d512_k128.svg)
+
+**Runtime dimensions:** `M`: Runtime number of token rows. `Nkv`: Number of latent vectors in the engine workspace.
+
+## sparse_mla_prefill_h64_d512_k640
 
-[Definition](../../definitions/sparse_mla/sparse_mla_prefill_h64_d512.json) · [Workloads](../../workloads/sparse_mla/sparse_mla_prefill_h64_d512.jsonl) · [SVG](sparse_mla_prefill_h64_d512.svg)
+[Definition](../../definitions/sparse_mla/sparse_mla_prefill_h64_d512_k640.json) · [Workloads](../../workloads/sparse_mla/sparse_mla_prefill_h64_d512_k640.jsonl) · [SVG](sparse_mla_prefill_h64_d512_k640.svg)
 
-![Shape distribution for sparse_mla_prefill_h64_d512: 30 workloads across 29 shapes](sparse_mla_prefill_h64_d512.svg)
+![Shape distribution for sparse_mla_prefill_h64_d512_k640: 20 workloads across 20 shapes](sparse_mla_prefill_h64_d512_k640.svg)
 
-**Runtime dimensions:** `M`: Runtime number of token rows. `Nkv`: Number of latent vectors in the engine workspace. `TopK`: Selected slots per query: 128 or 640.
+**Runtime dimensions:** `M`: Runtime number of token rows. `Nkv`: Number of latent vectors in the engine workspace.
diff --git a/docs/shape-distributions/dense_gemm_bf16_nt_n12288_k4096.svg b/docs/shape-distributions/dense_gemm_bf16_nt_n12288_k4096.svg
new file mode 100644
index 0000000000000000000000000000000000000000..2d4491ecb0190dc2162e1b94ba2a3442e1a0029e
--- /dev/null
+++ b/docs/shape-distributions/dense_gemm_bf16_nt_n12288_k4096.svg
@@ -0,0 +1,268 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="388.8pt" viewBox="0 0 849.6 388.8" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: dense_gemm_bf16_nt_n12288_k4096</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: dense_gemm_bf16_nt_n12288_k4096</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 388.8
+L 849.6 388.8
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+L 817.2 111.6
+L 57.6 111.6
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(106.895432 335.981156) rotate(-35)">4096</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_2"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(209.655172 335.981156) rotate(-35)">4125</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_3"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(312.414912 335.981156) rotate(-35)">6144</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_4"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(415.174653 335.981156) rotate(-35)">8192</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_5"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(512.722538 339.630536) rotate(-35)">12288</text>
+     </g>
+    </g>
+    <g id="xtick_6">
+     <g id="line2d_6"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(615.482278 339.630536) rotate(-35)">16384</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_7"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(718.242019 339.630536) rotate(-35)">24576</text>
+     </g>
+    </g>
+    <g id="text_8">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="437.4" y="361.196858" transform="rotate(-0 437.4 361.196858)">M — Runtime number of token rows</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_8">
+      <path d="M 57.6 308.16
+L 817.2 308.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_9"/>
+     <g id="text_9">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="311.958828" transform="rotate(-0 50.6 311.958828)">0</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_10">
+      <path d="M 57.6 270.826667
+L 817.2 270.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="274.625495" transform="rotate(-0 50.6 274.625495)">2</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_12">
+      <path d="M 57.6 233.493333
+L 817.2 233.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_13"/>
+     <g id="text_11">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="237.292161" transform="rotate(-0 50.6 237.292161)">4</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_14">
+      <path d="M 57.6 196.16
+L 817.2 196.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_15"/>
+     <g id="text_12">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="199.958828" transform="rotate(-0 50.6 199.958828)">6</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_16">
+      <path d="M 57.6 158.826667
+L 817.2 158.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_17"/>
+     <g id="text_13">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="162.625495" transform="rotate(-0 50.6 162.625495)">8</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_18">
+      <path d="M 57.6 121.493333
+L 817.2 121.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_19"/>
+     <g id="text_14">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="125.292161" transform="rotate(-0 50.6 125.292161)">10</text>
+     </g>
+    </g>
+    <g id="text_15">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="25.472656" y="209.88" transform="rotate(-90 25.472656 209.88)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 92.127273 308.16
+L 166.114286 308.16
+L 166.114286 140.16
+L 92.127273 140.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 194.887013 308.16
+L 268.874026 308.16
+L 268.874026 252.16
+L 194.887013 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 297.646753 308.16
+L 371.633766 308.16
+L 371.633766 252.16
+L 297.646753 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 400.406494 308.16
+L 474.393506 308.16
+L 474.393506 196.16
+L 400.406494 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 503.166234 308.16
+L 577.153247 308.16
+L 577.153247 270.826667
+L 503.166234 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 605.925974 308.16
+L 679.912987 308.16
+L 679.912987 196.16
+L 605.925974 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 708.685714 308.16
+L 782.672727 308.16
+L 782.672727 270.826667
+L 708.685714 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 57.6 308.16
+L 57.6 111.6
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="129.120779" y="133.997891" transform="rotate(-0 129.120779 133.997891)">9</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="231.880519" y="245.997891" transform="rotate(-0 231.880519 245.997891)">3</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="334.64026" y="245.997891" transform="rotate(-0 334.64026 245.997891)">3</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="437.4" y="189.997891" transform="rotate(-0 437.4 189.997891)">6</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="540.15974" y="264.664557" transform="rotate(-0 540.15974 264.664557)">2</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="642.919481" y="189.997891" transform="rotate(-0 642.919481 189.997891)">6</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="745.679221" y="264.664557" transform="rotate(-0 745.679221 264.664557)">2</text>
+   </g>
+  </g>
+  <g id="text_23">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_24">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">dense_gemm_bf16_nt_n12288_k4096</text>
+  </g>
+  <g id="text_25">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">31 captured workloads  |  7 distinct shapes</text>
+  </g>
+  <g id="text_26">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: N=12288, K=4096</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="381.6" transform="rotate(-0 29.736 381.6)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="p796238924c">
+   <rect x="57.6" y="111.6" width="759.6" height="196.56"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k12288.svg b/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k12288.svg
new file mode 100644
index 0000000000000000000000000000000000000000..26c8e61c96f66f3cf95899fe22e2f58154554d9c
--- /dev/null
+++ b/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k12288.svg
@@ -0,0 +1,268 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="388.8pt" viewBox="0 0 849.6 388.8" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: dense_gemm_bf16_nt_n4096_k12288</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: dense_gemm_bf16_nt_n4096_k12288</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 388.8
+L 849.6 388.8
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+L 817.2 111.6
+L 57.6 111.6
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(106.895432 335.981156) rotate(-35)">4096</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_2"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(209.655172 335.981156) rotate(-35)">4125</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_3"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(312.414912 335.981156) rotate(-35)">6144</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_4"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(415.174653 335.981156) rotate(-35)">8192</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_5"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(512.722538 339.630536) rotate(-35)">12288</text>
+     </g>
+    </g>
+    <g id="xtick_6">
+     <g id="line2d_6"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(615.482278 339.630536) rotate(-35)">16384</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_7"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(718.242019 339.630536) rotate(-35)">24576</text>
+     </g>
+    </g>
+    <g id="text_8">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="437.4" y="361.196858" transform="rotate(-0 437.4 361.196858)">M — Runtime number of token rows</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_8">
+      <path d="M 57.6 308.16
+L 817.2 308.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_9"/>
+     <g id="text_9">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="311.958828" transform="rotate(-0 50.6 311.958828)">0</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_10">
+      <path d="M 57.6 270.826667
+L 817.2 270.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="274.625495" transform="rotate(-0 50.6 274.625495)">2</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_12">
+      <path d="M 57.6 233.493333
+L 817.2 233.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_13"/>
+     <g id="text_11">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="237.292161" transform="rotate(-0 50.6 237.292161)">4</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_14">
+      <path d="M 57.6 196.16
+L 817.2 196.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_15"/>
+     <g id="text_12">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="199.958828" transform="rotate(-0 50.6 199.958828)">6</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_16">
+      <path d="M 57.6 158.826667
+L 817.2 158.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_17"/>
+     <g id="text_13">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="162.625495" transform="rotate(-0 50.6 162.625495)">8</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_18">
+      <path d="M 57.6 121.493333
+L 817.2 121.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_19"/>
+     <g id="text_14">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="125.292161" transform="rotate(-0 50.6 125.292161)">10</text>
+     </g>
+    </g>
+    <g id="text_15">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="25.472656" y="209.88" transform="rotate(-90 25.472656 209.88)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 92.127273 308.16
+L 166.114286 308.16
+L 166.114286 140.16
+L 92.127273 140.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 194.887013 308.16
+L 268.874026 308.16
+L 268.874026 252.16
+L 194.887013 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 297.646753 308.16
+L 371.633766 308.16
+L 371.633766 252.16
+L 297.646753 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 400.406494 308.16
+L 474.393506 308.16
+L 474.393506 196.16
+L 400.406494 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 503.166234 308.16
+L 577.153247 308.16
+L 577.153247 270.826667
+L 503.166234 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 605.925974 308.16
+L 679.912987 308.16
+L 679.912987 196.16
+L 605.925974 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 708.685714 308.16
+L 782.672727 308.16
+L 782.672727 270.826667
+L 708.685714 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 57.6 308.16
+L 57.6 111.6
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="129.120779" y="133.997891" transform="rotate(-0 129.120779 133.997891)">9</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="231.880519" y="245.997891" transform="rotate(-0 231.880519 245.997891)">3</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="334.64026" y="245.997891" transform="rotate(-0 334.64026 245.997891)">3</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="437.4" y="189.997891" transform="rotate(-0 437.4 189.997891)">6</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="540.15974" y="264.664557" transform="rotate(-0 540.15974 264.664557)">2</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="642.919481" y="189.997891" transform="rotate(-0 642.919481 189.997891)">6</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="745.679221" y="264.664557" transform="rotate(-0 745.679221 264.664557)">2</text>
+   </g>
+  </g>
+  <g id="text_23">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_24">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">dense_gemm_bf16_nt_n4096_k12288</text>
+  </g>
+  <g id="text_25">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">31 captured workloads  |  7 distinct shapes</text>
+  </g>
+  <g id="text_26">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: N=4096, K=12288</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="381.6" transform="rotate(-0 29.736 381.6)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="p796238924c">
+   <rect x="57.6" y="111.6" width="759.6" height="196.56"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k4096.svg b/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k4096.svg
new file mode 100644
index 0000000000000000000000000000000000000000..6b083b1e13b06331dd9218eb14093076f2583a0f
--- /dev/null
+++ b/docs/shape-distributions/dense_gemm_bf16_nt_n4096_k4096.svg
@@ -0,0 +1,268 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="388.8pt" viewBox="0 0 849.6 388.8" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: dense_gemm_bf16_nt_n4096_k4096</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: dense_gemm_bf16_nt_n4096_k4096</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 388.8
+L 849.6 388.8
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+L 817.2 111.6
+L 57.6 111.6
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(106.895432 335.981156) rotate(-35)">4096</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_2"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(209.655172 335.981156) rotate(-35)">4125</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_3"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(312.414912 335.981156) rotate(-35)">6144</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_4"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(415.174653 335.981156) rotate(-35)">8192</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_5"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(512.722538 339.630536) rotate(-35)">12288</text>
+     </g>
+    </g>
+    <g id="xtick_6">
+     <g id="line2d_6"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(615.482278 339.630536) rotate(-35)">16384</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_7"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(718.242019 339.630536) rotate(-35)">24576</text>
+     </g>
+    </g>
+    <g id="text_8">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="437.4" y="361.196858" transform="rotate(-0 437.4 361.196858)">M — Runtime number of token rows</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_8">
+      <path d="M 57.6 308.16
+L 817.2 308.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_9"/>
+     <g id="text_9">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="311.958828" transform="rotate(-0 50.6 311.958828)">0</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_10">
+      <path d="M 57.6 270.826667
+L 817.2 270.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="274.625495" transform="rotate(-0 50.6 274.625495)">2</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_12">
+      <path d="M 57.6 233.493333
+L 817.2 233.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_13"/>
+     <g id="text_11">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="237.292161" transform="rotate(-0 50.6 237.292161)">4</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_14">
+      <path d="M 57.6 196.16
+L 817.2 196.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_15"/>
+     <g id="text_12">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="199.958828" transform="rotate(-0 50.6 199.958828)">6</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_16">
+      <path d="M 57.6 158.826667
+L 817.2 158.826667
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_17"/>
+     <g id="text_13">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="162.625495" transform="rotate(-0 50.6 162.625495)">8</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_18">
+      <path d="M 57.6 121.493333
+L 817.2 121.493333
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_19"/>
+     <g id="text_14">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="125.292161" transform="rotate(-0 50.6 125.292161)">10</text>
+     </g>
+    </g>
+    <g id="text_15">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="25.472656" y="209.88" transform="rotate(-90 25.472656 209.88)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 92.127273 308.16
+L 166.114286 308.16
+L 166.114286 140.16
+L 92.127273 140.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 194.887013 308.16
+L 268.874026 308.16
+L 268.874026 252.16
+L 194.887013 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 297.646753 308.16
+L 371.633766 308.16
+L 371.633766 252.16
+L 297.646753 252.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 400.406494 308.16
+L 474.393506 308.16
+L 474.393506 196.16
+L 400.406494 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 503.166234 308.16
+L 577.153247 308.16
+L 577.153247 270.826667
+L 503.166234 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 605.925974 308.16
+L 679.912987 308.16
+L 679.912987 196.16
+L 605.925974 196.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 708.685714 308.16
+L 782.672727 308.16
+L 782.672727 270.826667
+L 708.685714 270.826667
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 57.6 308.16
+L 57.6 111.6
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="129.120779" y="133.997891" transform="rotate(-0 129.120779 133.997891)">9</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="231.880519" y="245.997891" transform="rotate(-0 231.880519 245.997891)">3</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="334.64026" y="245.997891" transform="rotate(-0 334.64026 245.997891)">3</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="437.4" y="189.997891" transform="rotate(-0 437.4 189.997891)">6</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="540.15974" y="264.664557" transform="rotate(-0 540.15974 264.664557)">2</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="642.919481" y="189.997891" transform="rotate(-0 642.919481 189.997891)">6</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="745.679221" y="264.664557" transform="rotate(-0 745.679221 264.664557)">2</text>
+   </g>
+  </g>
+  <g id="text_23">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_24">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">dense_gemm_bf16_nt_n4096_k4096</text>
+  </g>
+  <g id="text_25">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">31 captured workloads  |  7 distinct shapes</text>
+  </g>
+  <g id="text_26">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: N=4096, K=4096</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="381.6" transform="rotate(-0 29.736 381.6)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="p796238924c">
+   <rect x="57.6" y="111.6" width="759.6" height="196.56"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_g32.svg b/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_g32.svg
deleted file mode 100644
index b0e4b09f22db15cf633562bd12ed0b4caaf19779..0000000000000000000000000000000000000000
--- a/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_g32.svg
+++ /dev/null
@@ -1,545 +0,0 @@
-<?xml version="1.0" encoding="utf-8" standalone="no"?>
-<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
-  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="637.92pt" viewBox="0 0 849.6 637.92" xmlns="http://www.w3.org/2000/svg" version="1.1">
- <title>Shape distribution: grouped_gemm_fp8_fp4_e384_g32</title>
- <metadata>
-  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
-   <cc:Work>
-    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
-    <dc:title>Shape distribution: grouped_gemm_fp8_fp4_e384_g32</dc:title>
-    <dc:format>image/svg+xml</dc:format>
-    <dc:creator>
-     <cc:Agent>
-      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
-     </cc:Agent>
-    </dc:creator>
-   </cc:Work>
-  </rdf:RDF>
- </metadata>
- <defs>
-  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
- </defs>
- <g id="figure_1">
-  <g id="patch_1">
-   <path d="M 0 637.92
-L 849.6 637.92
-L 849.6 0
-L 0 0
-z
-" style="fill: #ffffff"/>
-  </g>
-  <g id="axes_1">
-   <g id="patch_2">
-    <path d="M 184.32 573.12
-L 817.2 573.12
-L 817.2 133.2
-L 184.32 133.2
-z
-" style="fill: #ffffff"/>
-   </g>
-   <g id="matplotlib.axis_1">
-    <g id="xtick_1">
-     <g id="line2d_1">
-      <path d="M 184.32 573.12
-L 184.32 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_2"/>
-     <g id="text_1">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="184.32" y="587.717656" transform="rotate(-0 184.32 587.717656)">0</text>
-     </g>
-    </g>
-    <g id="xtick_2">
-     <g id="line2d_3">
-      <path d="M 266.683352 573.12
-L 266.683352 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_4"/>
-     <g id="text_2">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="266.683352" y="587.717656" transform="rotate(-0 266.683352 587.717656)">5</text>
-     </g>
-    </g>
-    <g id="xtick_3">
-     <g id="line2d_5">
-      <path d="M 349.046705 573.12
-L 349.046705 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_6"/>
-     <g id="text_3">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="349.046705" y="587.717656" transform="rotate(-0 349.046705 587.717656)">10</text>
-     </g>
-    </g>
-    <g id="xtick_4">
-     <g id="line2d_7">
-      <path d="M 431.410057 573.12
-L 431.410057 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_8"/>
-     <g id="text_4">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="431.410057" y="587.717656" transform="rotate(-0 431.410057 587.717656)">15</text>
-     </g>
-    </g>
-    <g id="xtick_5">
-     <g id="line2d_9">
-      <path d="M 513.77341 573.12
-L 513.77341 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_10"/>
-     <g id="text_5">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="513.77341" y="587.717656" transform="rotate(-0 513.77341 587.717656)">20</text>
-     </g>
-    </g>
-    <g id="xtick_6">
-     <g id="line2d_11">
-      <path d="M 596.136762 573.12
-L 596.136762 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_12"/>
-     <g id="text_6">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="596.136762" y="587.717656" transform="rotate(-0 596.136762 587.717656)">25</text>
-     </g>
-    </g>
-    <g id="xtick_7">
-     <g id="line2d_13">
-      <path d="M 678.500115 573.12
-L 678.500115 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_14"/>
-     <g id="text_7">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="678.500115" y="587.717656" transform="rotate(-0 678.500115 587.717656)">30</text>
-     </g>
-    </g>
-    <g id="xtick_8">
-     <g id="line2d_15">
-      <path d="M 760.863467 573.12
-L 760.863467 133.2
-" clip-path="url(#pfce6301e93)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_16"/>
-     <g id="text_8">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="760.863467" y="587.717656" transform="rotate(-0 760.863467 587.717656)">35</text>
-     </g>
-    </g>
-    <g id="text_9">
-     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="500.76" y="607.718437" transform="rotate(-0 500.76 607.718437)">Captured workloads (count)</text>
-    </g>
-   </g>
-   <g id="matplotlib.axis_2">
-    <g id="ytick_1">
-     <g id="line2d_17"/>
-     <g id="text_10">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="162.887231" transform="rotate(-0 177.32 162.887231)">(6, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_2">
-     <g id="line2d_18"/>
-     <g id="text_11">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="181.334061" transform="rotate(-0 177.32 181.334061)">(6, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_3">
-     <g id="line2d_19"/>
-     <g id="text_12">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="199.780891" transform="rotate(-0 177.32 199.780891)">(12, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_4">
-     <g id="line2d_20"/>
-     <g id="text_13">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="218.227721" transform="rotate(-0 177.32 218.227721)">(12, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_5">
-     <g id="line2d_21"/>
-     <g id="text_14">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="236.674551" transform="rotate(-0 177.32 236.674551)">(18, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_6">
-     <g id="line2d_22"/>
-     <g id="text_15">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="255.121381" transform="rotate(-0 177.32 255.121381)">(18, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_7">
-     <g id="line2d_23"/>
-     <g id="text_16">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="273.568211" transform="rotate(-0 177.32 273.568211)">(768, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_8">
-     <g id="line2d_24"/>
-     <g id="text_17">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="292.015041" transform="rotate(-0 177.32 292.015041)">(768, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_9">
-     <g id="line2d_25"/>
-     <g id="text_18">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="310.461871" transform="rotate(-0 177.32 310.461871)">(2688, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_10">
-     <g id="line2d_26"/>
-     <g id="text_19">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="328.9087" transform="rotate(-0 177.32 328.9087)">(2688, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_11">
-     <g id="line2d_27"/>
-     <g id="text_20">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="347.35553" transform="rotate(-0 177.32 347.35553)">(11904, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_12">
-     <g id="line2d_28"/>
-     <g id="text_21">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="365.80236" transform="rotate(-0 177.32 365.80236)">(11904, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_13">
-     <g id="line2d_29"/>
-     <g id="text_22">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="384.24919" transform="rotate(-0 177.32 384.24919)">(14592, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_14">
-     <g id="line2d_30"/>
-     <g id="text_23">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="402.69602" transform="rotate(-0 177.32 402.69602)">(14592, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_15">
-     <g id="line2d_31"/>
-     <g id="text_24">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="421.14285" transform="rotate(-0 177.32 421.14285)">(17664, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_16">
-     <g id="line2d_32"/>
-     <g id="text_25">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="439.58968" transform="rotate(-0 177.32 439.58968)">(17664, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_17">
-     <g id="line2d_33"/>
-     <g id="text_26">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="458.03651" transform="rotate(-0 177.32 458.03651)">(20736, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_18">
-     <g id="line2d_34"/>
-     <g id="text_27">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="476.48334" transform="rotate(-0 177.32 476.48334)">(20736, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_19">
-     <g id="line2d_35"/>
-     <g id="text_28">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="494.93017" transform="rotate(-0 177.32 494.93017)">(24192, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_20">
-     <g id="line2d_36"/>
-     <g id="text_29">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="513.377" transform="rotate(-0 177.32 513.377)">(24192, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-    <g id="ytick_21">
-     <g id="line2d_37"/>
-     <g id="text_30">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="531.82383" transform="rotate(-0 177.32 531.82383)">(24576, 4608, 5120, 2560, 160)</text>
-     </g>
-    </g>
-    <g id="ytick_22">
-     <g id="line2d_38"/>
-     <g id="text_31">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="177.32" y="550.27066" transform="rotate(-0 177.32 550.27066)">(24576, 5120, 2304, 1152, 72)</text>
-     </g>
-    </g>
-   </g>
-   <g id="patch_3">
-    <path d="M 184.32 153.196364
-L 744.390796 153.196364
-L 744.390796 165.740208
-L 184.32 165.740208
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_4">
-    <path d="M 184.32 171.643194
-L 744.390796 171.643194
-L 744.390796 184.187038
-L 184.32 184.187038
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_5">
-    <path d="M 184.32 190.090023
-L 250.210682 190.090023
-L 250.210682 202.633868
-L 184.32 202.633868
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_6">
-    <path d="M 184.32 208.536853
-L 250.210682 208.536853
-L 250.210682 221.080698
-L 184.32 221.080698
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_7">
-    <path d="M 184.32 226.983683
-L 250.210682 226.983683
-L 250.210682 239.527528
-L 184.32 239.527528
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_8">
-    <path d="M 184.32 245.430513
-L 250.210682 245.430513
-L 250.210682 257.974358
-L 184.32 257.974358
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_9">
-    <path d="M 184.32 263.877343
-L 200.79267 263.877343
-L 200.79267 276.421188
-L 184.32 276.421188
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_10">
-    <path d="M 184.32 282.324173
-L 200.79267 282.324173
-L 200.79267 294.868017
-L 184.32 294.868017
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_11">
-    <path d="M 184.32 300.771003
-L 200.79267 300.771003
-L 200.79267 313.314847
-L 184.32 313.314847
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_12">
-    <path d="M 184.32 319.217833
-L 200.79267 319.217833
-L 200.79267 331.761677
-L 184.32 331.761677
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_13">
-    <path d="M 184.32 337.664663
-L 200.79267 337.664663
-L 200.79267 350.208507
-L 184.32 350.208507
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_14">
-    <path d="M 184.32 356.111493
-L 200.79267 356.111493
-L 200.79267 368.655337
-L 184.32 368.655337
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_15">
-    <path d="M 184.32 374.558323
-L 200.79267 374.558323
-L 200.79267 387.102167
-L 184.32 387.102167
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_16">
-    <path d="M 184.32 393.005153
-L 200.79267 393.005153
-L 200.79267 405.548997
-L 184.32 405.548997
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_17">
-    <path d="M 184.32 411.451983
-L 200.79267 411.451983
-L 200.79267 423.995827
-L 184.32 423.995827
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_18">
-    <path d="M 184.32 429.898812
-L 200.79267 429.898812
-L 200.79267 442.442657
-L 184.32 442.442657
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_19">
-    <path d="M 184.32 448.345642
-L 200.79267 448.345642
-L 200.79267 460.889487
-L 184.32 460.889487
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_20">
-    <path d="M 184.32 466.792472
-L 200.79267 466.792472
-L 200.79267 479.336317
-L 184.32 479.336317
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_21">
-    <path d="M 184.32 485.239302
-L 200.79267 485.239302
-L 200.79267 497.783147
-L 184.32 497.783147
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_22">
-    <path d="M 184.32 503.686132
-L 200.79267 503.686132
-L 200.79267 516.229977
-L 184.32 516.229977
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_23">
-    <path d="M 184.32 522.132962
-L 233.738011 522.132962
-L 233.738011 534.676806
-L 184.32 534.676806
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_24">
-    <path d="M 184.32 540.579792
-L 233.738011 540.579792
-L 233.738011 553.123636
-L 184.32 553.123636
-z
-" clip-path="url(#pfce6301e93)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_25">
-    <path d="M 184.32 573.12
-L 184.32 133.2
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="patch_26">
-    <path d="M 184.32 573.12
-L 817.2 573.12
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="text_32">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="748.390796" y="161.806176" transform="rotate(-0 748.390796 161.806176)">34</text>
-   </g>
-   <g id="text_33">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="748.390796" y="180.253006" transform="rotate(-0 748.390796 180.253006)">34</text>
-   </g>
-   <g id="text_34">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="254.210682" y="198.699836" transform="rotate(-0 254.210682 198.699836)">4</text>
-   </g>
-   <g id="text_35">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="254.210682" y="217.146666" transform="rotate(-0 254.210682 217.146666)">4</text>
-   </g>
-   <g id="text_36">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="254.210682" y="235.593496" transform="rotate(-0 254.210682 235.593496)">4</text>
-   </g>
-   <g id="text_37">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="254.210682" y="254.040326" transform="rotate(-0 254.210682 254.040326)">4</text>
-   </g>
-   <g id="text_38">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="272.487156" transform="rotate(-0 204.79267 272.487156)">1</text>
-   </g>
-   <g id="text_39">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="290.933986" transform="rotate(-0 204.79267 290.933986)">1</text>
-   </g>
-   <g id="text_40">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="309.380816" transform="rotate(-0 204.79267 309.380816)">1</text>
-   </g>
-   <g id="text_41">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="327.827646" transform="rotate(-0 204.79267 327.827646)">1</text>
-   </g>
-   <g id="text_42">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="346.274476" transform="rotate(-0 204.79267 346.274476)">1</text>
-   </g>
-   <g id="text_43">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="364.721306" transform="rotate(-0 204.79267 364.721306)">1</text>
-   </g>
-   <g id="text_44">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="383.168136" transform="rotate(-0 204.79267 383.168136)">1</text>
-   </g>
-   <g id="text_45">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="401.614965" transform="rotate(-0 204.79267 401.614965)">1</text>
-   </g>
-   <g id="text_46">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="420.061795" transform="rotate(-0 204.79267 420.061795)">1</text>
-   </g>
-   <g id="text_47">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="438.508625" transform="rotate(-0 204.79267 438.508625)">1</text>
-   </g>
-   <g id="text_48">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="456.955455" transform="rotate(-0 204.79267 456.955455)">1</text>
-   </g>
-   <g id="text_49">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="475.402285" transform="rotate(-0 204.79267 475.402285)">1</text>
-   </g>
-   <g id="text_50">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="493.849115" transform="rotate(-0 204.79267 493.849115)">1</text>
-   </g>
-   <g id="text_51">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="204.79267" y="512.295945" transform="rotate(-0 204.79267 512.295945)">1</text>
-   </g>
-   <g id="text_52">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="237.738011" y="530.742775" transform="rotate(-0 237.738011 530.742775)">3</text>
-   </g>
-   <g id="text_53">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="237.738011" y="549.189605" transform="rotate(-0 237.738011 549.189605)">3</text>
-   </g>
-  </g>
-  <g id="text_54">
-   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
-  </g>
-  <g id="text_55">
-   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">grouped_gemm_fp8_fp4_e384_g32</text>
-  </g>
-  <g id="text_56">
-   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">104 captured workloads  |  22 distinct shapes</text>
-  </g>
-  <g id="text_57">
-   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: E=384, group_size=32</text>
-  </g>
-  <g id="text_58">
-   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="184.32" y="111.6" transform="rotate(-0 184.32 111.6)">Runtime shape: (M, N, K, K_packed, K_groups)</text>
-  </g>
-  <g id="text_59">
-   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="630.72" transform="rotate(-0 29.736 630.72)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
-  </g>
- </g>
- <defs>
-  <clipPath id="pfce6301e93">
-   <rect x="184.32" y="133.2" width="632.88" height="439.92"/>
-  </clipPath>
- </defs>
-</svg>
diff --git a/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.svg b/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.svg
new file mode 100644
index 0000000000000000000000000000000000000000..6200fb4821719c02302d20145cd9c93f010125e7
--- /dev/null
+++ b/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.svg
@@ -0,0 +1,347 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="388.8pt" viewBox="0 0 849.6 388.8" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: grouped_gemm_fp8_fp4_e384_n4608_k5120_g32</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: grouped_gemm_fp8_fp4_e384_n4608_k5120_g32</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 388.8
+L 849.6 388.8
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+L 817.2 111.6
+L 57.6 111.6
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(108.727449 325.033016) rotate(-35)">6</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_2"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(167.932148 328.682396) rotate(-35)">12</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_3"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(232.348702 328.682396) rotate(-35)">18</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_4"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(291.5534 332.331776) rotate(-35)">768</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_5"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(350.758099 335.981156) rotate(-35)">2688</text>
+     </g>
+    </g>
+    <g id="xtick_6">
+     <g id="line2d_6"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(409.962798 339.630536) rotate(-35)">11904</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_7"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(474.379351 339.630536) rotate(-35)">14592</text>
+     </g>
+    </g>
+    <g id="xtick_8">
+     <g id="line2d_8"/>
+     <g id="text_8">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(538.795905 339.630536) rotate(-35)">17664</text>
+     </g>
+    </g>
+    <g id="xtick_9">
+     <g id="line2d_9"/>
+     <g id="text_9">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(603.212459 339.630536) rotate(-35)">20736</text>
+     </g>
+    </g>
+    <g id="xtick_10">
+     <g id="line2d_10"/>
+     <g id="text_10">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(667.629012 339.630536) rotate(-35)">24192</text>
+     </g>
+    </g>
+    <g id="xtick_11">
+     <g id="line2d_11"/>
+     <g id="text_11">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(732.045566 339.630536) rotate(-35)">24576</text>
+     </g>
+    </g>
+    <g id="text_12">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="437.4" y="361.196858" transform="rotate(-0 437.4 361.196858)">M — Runtime number of token rows</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_12">
+      <path d="M 57.6 308.16
+L 817.2 308.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_13"/>
+     <g id="text_13">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="311.958828" transform="rotate(-0 50.6 311.958828)">0</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_14">
+      <path d="M 57.6 278.512941
+L 817.2 278.512941
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_15"/>
+     <g id="text_14">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="282.311769" transform="rotate(-0 50.6 282.311769)">6</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_16">
+      <path d="M 57.6 248.865882
+L 817.2 248.865882
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_17"/>
+     <g id="text_15">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="252.66471" transform="rotate(-0 50.6 252.66471)">12</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_18">
+      <path d="M 57.6 219.218824
+L 817.2 219.218824
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_19"/>
+     <g id="text_16">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="223.017652" transform="rotate(-0 50.6 223.017652)">18</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_20">
+      <path d="M 57.6 189.571765
+L 817.2 189.571765
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_21"/>
+     <g id="text_17">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="193.370593" transform="rotate(-0 50.6 193.370593)">24</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_22">
+      <path d="M 57.6 159.924706
+L 817.2 159.924706
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_23"/>
+     <g id="text_18">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="163.723534" transform="rotate(-0 50.6 163.723534)">30</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_24">
+      <path d="M 57.6 130.277647
+L 817.2 130.277647
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_25"/>
+     <g id="text_19">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="134.076475" transform="rotate(-0 50.6 134.076475)">36</text>
+     </g>
+    </g>
+    <g id="text_20">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="25.472656" y="209.88" transform="rotate(-90 25.472656 209.88)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 92.127273 308.16
+L 138.507191 308.16
+L 138.507191 140.16
+L 92.127273 140.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 156.543826 308.16
+L 202.923745 308.16
+L 202.923745 288.395294
+L 156.543826 288.395294
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 220.96038 308.16
+L 267.340299 308.16
+L 267.340299 288.395294
+L 220.96038 288.395294
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 285.376934 308.16
+L 331.756852 308.16
+L 331.756852 303.218824
+L 285.376934 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 349.793487 308.16
+L 396.173406 308.16
+L 396.173406 303.218824
+L 349.793487 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 414.210041 308.16
+L 460.589959 308.16
+L 460.589959 303.218824
+L 414.210041 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 478.626594 308.16
+L 525.006513 308.16
+L 525.006513 303.218824
+L 478.626594 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 543.043148 308.16
+L 589.423066 308.16
+L 589.423066 303.218824
+L 543.043148 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 607.459701 308.16
+L 653.83962 308.16
+L 653.83962 303.218824
+L 607.459701 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 671.876255 308.16
+L 718.256174 308.16
+L 718.256174 303.218824
+L 671.876255 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 736.292809 308.16
+L 782.672727 308.16
+L 782.672727 293.336471
+L 736.292809 293.336471
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_14">
+    <path d="M 57.6 308.16
+L 57.6 111.6
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_15">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="115.317232" y="133.997891" transform="rotate(-0 115.317232 133.997891)">34</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="179.733786" y="282.233185" transform="rotate(-0 179.733786 282.233185)">4</text>
+   </g>
+   <g id="text_23">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="244.150339" y="282.233185" transform="rotate(-0 244.150339 282.233185)">4</text>
+   </g>
+   <g id="text_24">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="308.566893" y="297.056714" transform="rotate(-0 308.566893 297.056714)">1</text>
+   </g>
+   <g id="text_25">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="372.983446" y="297.056714" transform="rotate(-0 372.983446 297.056714)">1</text>
+   </g>
+   <g id="text_26">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="437.4" y="297.056714" transform="rotate(-0 437.4 297.056714)">1</text>
+   </g>
+   <g id="text_27">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="501.816554" y="297.056714" transform="rotate(-0 501.816554 297.056714)">1</text>
+   </g>
+   <g id="text_28">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="566.233107" y="297.056714" transform="rotate(-0 566.233107 297.056714)">1</text>
+   </g>
+   <g id="text_29">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="630.649661" y="297.056714" transform="rotate(-0 630.649661 297.056714)">1</text>
+   </g>
+   <g id="text_30">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="695.066214" y="297.056714" transform="rotate(-0 695.066214 297.056714)">1</text>
+   </g>
+   <g id="text_31">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="759.482768" y="287.174361" transform="rotate(-0 759.482768 287.174361)">3</text>
+   </g>
+  </g>
+  <g id="text_32">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_33">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">grouped_gemm_fp8_fp4_e384_n4608_k5120_g32</text>
+  </g>
+  <g id="text_34">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">52 captured workloads  |  11 distinct shapes</text>
+  </g>
+  <g id="text_35">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: E=384, N=4608, K=5120, K_packed=2560, K_groups=160, group_size=32</text>
+  </g>
+  <g id="text_36">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="381.6" transform="rotate(-0 29.736 381.6)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="p796238924c">
+   <rect x="57.6" y="111.6" width="759.6" height="196.56"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.svg b/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.svg
new file mode 100644
index 0000000000000000000000000000000000000000..58dd18613a525704f5838103563be5cdf674c890
--- /dev/null
+++ b/docs/shape-distributions/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.svg
@@ -0,0 +1,347 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="388.8pt" viewBox="0 0 849.6 388.8" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: grouped_gemm_fp8_fp4_e384_n5120_k2304_g32</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: grouped_gemm_fp8_fp4_e384_n5120_k2304_g32</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 388.8
+L 849.6 388.8
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+L 817.2 111.6
+L 57.6 111.6
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(108.727449 325.033016) rotate(-35)">6</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_2"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(167.932148 328.682396) rotate(-35)">12</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_3"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(232.348702 328.682396) rotate(-35)">18</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_4"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(291.5534 332.331776) rotate(-35)">768</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_5"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(350.758099 335.981156) rotate(-35)">2688</text>
+     </g>
+    </g>
+    <g id="xtick_6">
+     <g id="line2d_6"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(409.962798 339.630536) rotate(-35)">11904</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_7"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(474.379351 339.630536) rotate(-35)">14592</text>
+     </g>
+    </g>
+    <g id="xtick_8">
+     <g id="line2d_8"/>
+     <g id="text_8">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(538.795905 339.630536) rotate(-35)">17664</text>
+     </g>
+    </g>
+    <g id="xtick_9">
+     <g id="line2d_9"/>
+     <g id="text_9">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(603.212459 339.630536) rotate(-35)">20736</text>
+     </g>
+    </g>
+    <g id="xtick_10">
+     <g id="line2d_10"/>
+     <g id="text_10">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(667.629012 339.630536) rotate(-35)">24192</text>
+     </g>
+    </g>
+    <g id="xtick_11">
+     <g id="line2d_11"/>
+     <g id="text_11">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; fill: #334155" transform="translate(732.045566 339.630536) rotate(-35)">24576</text>
+     </g>
+    </g>
+    <g id="text_12">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="437.4" y="361.196858" transform="rotate(-0 437.4 361.196858)">M — Runtime number of token rows</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_12">
+      <path d="M 57.6 308.16
+L 817.2 308.16
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_13"/>
+     <g id="text_13">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="311.958828" transform="rotate(-0 50.6 311.958828)">0</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_14">
+      <path d="M 57.6 278.512941
+L 817.2 278.512941
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_15"/>
+     <g id="text_14">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="282.311769" transform="rotate(-0 50.6 282.311769)">6</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_16">
+      <path d="M 57.6 248.865882
+L 817.2 248.865882
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_17"/>
+     <g id="text_15">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="252.66471" transform="rotate(-0 50.6 252.66471)">12</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_18">
+      <path d="M 57.6 219.218824
+L 817.2 219.218824
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_19"/>
+     <g id="text_16">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="223.017652" transform="rotate(-0 50.6 223.017652)">18</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_20">
+      <path d="M 57.6 189.571765
+L 817.2 189.571765
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_21"/>
+     <g id="text_17">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="193.370593" transform="rotate(-0 50.6 193.370593)">24</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_22">
+      <path d="M 57.6 159.924706
+L 817.2 159.924706
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_23"/>
+     <g id="text_18">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="163.723534" transform="rotate(-0 50.6 163.723534)">30</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_24">
+      <path d="M 57.6 130.277647
+L 817.2 130.277647
+" clip-path="url(#p796238924c)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_25"/>
+     <g id="text_19">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: end; fill: #334155" x="50.6" y="134.076475" transform="rotate(-0 50.6 134.076475)">36</text>
+     </g>
+    </g>
+    <g id="text_20">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="25.472656" y="209.88" transform="rotate(-90 25.472656 209.88)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 92.127273 308.16
+L 138.507191 308.16
+L 138.507191 140.16
+L 92.127273 140.16
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 156.543826 308.16
+L 202.923745 308.16
+L 202.923745 288.395294
+L 156.543826 288.395294
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 220.96038 308.16
+L 267.340299 308.16
+L 267.340299 288.395294
+L 220.96038 288.395294
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 285.376934 308.16
+L 331.756852 308.16
+L 331.756852 303.218824
+L 285.376934 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 349.793487 308.16
+L 396.173406 308.16
+L 396.173406 303.218824
+L 349.793487 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 414.210041 308.16
+L 460.589959 308.16
+L 460.589959 303.218824
+L 414.210041 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 478.626594 308.16
+L 525.006513 308.16
+L 525.006513 303.218824
+L 478.626594 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 543.043148 308.16
+L 589.423066 308.16
+L 589.423066 303.218824
+L 543.043148 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 607.459701 308.16
+L 653.83962 308.16
+L 653.83962 303.218824
+L 607.459701 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 671.876255 308.16
+L 718.256174 308.16
+L 718.256174 303.218824
+L 671.876255 303.218824
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 736.292809 308.16
+L 782.672727 308.16
+L 782.672727 293.336471
+L 736.292809 293.336471
+z
+" clip-path="url(#p796238924c)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_14">
+    <path d="M 57.6 308.16
+L 57.6 111.6
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_15">
+    <path d="M 57.6 308.16
+L 817.2 308.16
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="115.317232" y="133.997891" transform="rotate(-0 115.317232 133.997891)">34</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="179.733786" y="282.233185" transform="rotate(-0 179.733786 282.233185)">4</text>
+   </g>
+   <g id="text_23">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="244.150339" y="282.233185" transform="rotate(-0 244.150339 282.233185)">4</text>
+   </g>
+   <g id="text_24">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="308.566893" y="297.056714" transform="rotate(-0 308.566893 297.056714)">1</text>
+   </g>
+   <g id="text_25">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="372.983446" y="297.056714" transform="rotate(-0 372.983446 297.056714)">1</text>
+   </g>
+   <g id="text_26">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="437.4" y="297.056714" transform="rotate(-0 437.4 297.056714)">1</text>
+   </g>
+   <g id="text_27">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="501.816554" y="297.056714" transform="rotate(-0 501.816554 297.056714)">1</text>
+   </g>
+   <g id="text_28">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="566.233107" y="297.056714" transform="rotate(-0 566.233107 297.056714)">1</text>
+   </g>
+   <g id="text_29">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="630.649661" y="297.056714" transform="rotate(-0 630.649661 297.056714)">1</text>
+   </g>
+   <g id="text_30">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="695.066214" y="297.056714" transform="rotate(-0 695.066214 297.056714)">1</text>
+   </g>
+   <g id="text_31">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #243c52" x="759.482768" y="287.174361" transform="rotate(-0 759.482768 287.174361)">3</text>
+   </g>
+  </g>
+  <g id="text_32">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_33">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">grouped_gemm_fp8_fp4_e384_n5120_k2304_g32</text>
+  </g>
+  <g id="text_34">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">52 captured workloads  |  11 distinct shapes</text>
+  </g>
+  <g id="text_35">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: E=384, N=5120, K=2304, K_packed=1152, K_groups=72, group_size=32</text>
+  </g>
+  <g id="text_36">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="381.6" transform="rotate(-0 29.736 381.6)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="p796238924c">
+   <rect x="57.6" y="111.6" width="759.6" height="196.56"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/neighborhood_attention_3d_d64.svg b/docs/shape-distributions/neighborhood_attention_3d_d64.svg
deleted file mode 100644
index 1d1f628d4d41500cd80ec5c0080632a5fef67dc1..0000000000000000000000000000000000000000
--- a/docs/shape-distributions/neighborhood_attention_3d_d64.svg
+++ /dev/null
@@ -1,870 +0,0 @@
-<?xml version="1.0" encoding="utf-8" standalone="no"?>
-<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
-  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="1101.6pt" viewBox="0 0 849.6 1101.6" xmlns="http://www.w3.org/2000/svg" version="1.1">
- <title>Shape distribution: neighborhood_attention_3d_d64</title>
- <metadata>
-  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
-   <cc:Work>
-    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
-    <dc:title>Shape distribution: neighborhood_attention_3d_d64</dc:title>
-    <dc:format>image/svg+xml</dc:format>
-    <dc:creator>
-     <cc:Agent>
-      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
-     </cc:Agent>
-    </dc:creator>
-   </cc:Work>
-  </rdf:RDF>
- </metadata>
- <defs>
-  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
- </defs>
- <g id="figure_1">
-  <g id="patch_1">
-   <path d="M 0 1101.6
-L 849.6 1101.6
-L 849.6 0
-L 0 0
-z
-" style="fill: #ffffff"/>
-  </g>
-  <g id="axes_1">
-   <g id="patch_2">
-    <path d="M 144 1036.8
-L 817.2 1036.8
-L 817.2 133.2
-L 144 133.2
-z
-" style="fill: #ffffff"/>
-   </g>
-   <g id="matplotlib.axis_1">
-    <g id="xtick_1">
-     <g id="line2d_1">
-      <path d="M 144 1036.8
-L 144 133.2
-" clip-path="url(#p03b2f57e79)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_2"/>
-     <g id="text_1">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="1051.397656" transform="rotate(-0 144 1051.397656)">0</text>
-     </g>
-    </g>
-    <g id="xtick_2">
-     <g id="line2d_3">
-      <path d="M 739.752212 1036.8
-L 739.752212 133.2
-" clip-path="url(#p03b2f57e79)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_4"/>
-     <g id="text_2">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="1051.397656" transform="rotate(-0 739.752212 1051.397656)">1</text>
-     </g>
-    </g>
-    <g id="text_3">
-     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="1071.398438" transform="rotate(-0 480.6 1071.398438)">Captured workloads (count)</text>
-    </g>
-   </g>
-   <g id="matplotlib.axis_2">
-    <g id="ytick_1">
-     <g id="line2d_5"/>
-     <g id="text_4">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="183.94267" transform="rotate(-0 137 183.94267)">(1, 9, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_2">
-     <g id="line2d_6"/>
-     <g id="text_5">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="202.327955" transform="rotate(-0 137 202.327955)">(1, 9, 14, 24, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_3">
-     <g id="line2d_7"/>
-     <g id="text_6">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="220.71324" transform="rotate(-0 137 220.71324)">(1, 9, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_4">
-     <g id="line2d_8"/>
-     <g id="text_7">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="239.098525" transform="rotate(-0 137 239.098525)">(1, 9, 28, 48, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_5">
-     <g id="line2d_9"/>
-     <g id="text_8">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="257.483811" transform="rotate(-0 137 257.483811)">(1, 17, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_6">
-     <g id="line2d_10"/>
-     <g id="text_9">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="275.869096" transform="rotate(-0 137 275.869096)">(1, 17, 28, 48, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_7">
-     <g id="line2d_11"/>
-     <g id="text_10">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="294.254381" transform="rotate(-0 137 294.254381)">(1, 18, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_8">
-     <g id="line2d_12"/>
-     <g id="text_11">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="312.639666" transform="rotate(-0 137 312.639666)">(1, 18, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_9">
-     <g id="line2d_13"/>
-     <g id="text_12">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="331.024952" transform="rotate(-0 137 331.024952)">(1, 33, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_10">
-     <g id="line2d_14"/>
-     <g id="text_13">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="349.410237" transform="rotate(-0 137 349.410237)">(1, 33, 56, 96, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_11">
-     <g id="line2d_15"/>
-     <g id="text_14">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="367.795522" transform="rotate(-0 137 367.795522)">(1, 35, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_12">
-     <g id="line2d_16"/>
-     <g id="text_15">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="386.180807" transform="rotate(-0 137 386.180807)">(1, 49, 80, 128, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_13">
-     <g id="line2d_17"/>
-     <g id="text_16">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="404.566093" transform="rotate(-0 137 404.566093)">(1, 49, 112, 192, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_14">
-     <g id="line2d_18"/>
-     <g id="text_17">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="422.951378" transform="rotate(-0 137 422.951378)">(1, 69, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_15">
-     <g id="line2d_19"/>
-     <g id="text_18">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="441.336663" transform="rotate(-0 137 441.336663)">(1, 121, 80, 128, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_16">
-     <g id="line2d_20"/>
-     <g id="text_19">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="459.721948" transform="rotate(-0 137 459.721948)">(2, 9, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_17">
-     <g id="line2d_21"/>
-     <g id="text_20">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="478.107234" transform="rotate(-0 137 478.107234)">(2, 9, 14, 24, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_18">
-     <g id="line2d_22"/>
-     <g id="text_21">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="496.492519" transform="rotate(-0 137 496.492519)">(2, 9, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_19">
-     <g id="line2d_23"/>
-     <g id="text_22">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="514.877804" transform="rotate(-0 137 514.877804)">(2, 9, 28, 48, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_20">
-     <g id="line2d_24"/>
-     <g id="text_23">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="533.26309" transform="rotate(-0 137 533.26309)">(2, 17, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_21">
-     <g id="line2d_25"/>
-     <g id="text_24">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="551.648375" transform="rotate(-0 137 551.648375)">(2, 17, 28, 48, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_22">
-     <g id="line2d_26"/>
-     <g id="text_25">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="570.03366" transform="rotate(-0 137 570.03366)">(2, 18, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_23">
-     <g id="line2d_27"/>
-     <g id="text_26">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="588.418945" transform="rotate(-0 137 588.418945)">(2, 18, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_24">
-     <g id="line2d_28"/>
-     <g id="text_27">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="606.804231" transform="rotate(-0 137 606.804231)">(2, 33, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_25">
-     <g id="line2d_29"/>
-     <g id="text_28">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="625.189516" transform="rotate(-0 137 625.189516)">(2, 33, 56, 96, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_26">
-     <g id="line2d_30"/>
-     <g id="text_29">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="643.574801" transform="rotate(-0 137 643.574801)">(2, 35, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_27">
-     <g id="line2d_31"/>
-     <g id="text_30">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="661.960086" transform="rotate(-0 137 661.960086)">(2, 49, 80, 128, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_28">
-     <g id="line2d_32"/>
-     <g id="text_31">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="680.345372" transform="rotate(-0 137 680.345372)">(2, 49, 112, 192, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_29">
-     <g id="line2d_33"/>
-     <g id="text_32">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="698.730657" transform="rotate(-0 137 698.730657)">(2, 69, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_30">
-     <g id="line2d_34"/>
-     <g id="text_33">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="717.115942" transform="rotate(-0 137 717.115942)">(2, 121, 80, 128, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_31">
-     <g id="line2d_35"/>
-     <g id="text_34">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="735.501227" transform="rotate(-0 137 735.501227)">(4, 9, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_32">
-     <g id="line2d_36"/>
-     <g id="text_35">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="753.886513" transform="rotate(-0 137 753.886513)">(4, 9, 14, 24, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_33">
-     <g id="line2d_37"/>
-     <g id="text_36">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="772.271798" transform="rotate(-0 137 772.271798)">(4, 9, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_34">
-     <g id="line2d_38"/>
-     <g id="text_37">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="790.657083" transform="rotate(-0 137 790.657083)">(4, 9, 28, 48, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_35">
-     <g id="line2d_39"/>
-     <g id="text_38">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="809.042368" transform="rotate(-0 137 809.042368)">(4, 17, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_36">
-     <g id="line2d_40"/>
-     <g id="text_39">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="827.427654" transform="rotate(-0 137 827.427654)">(4, 17, 28, 48, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_37">
-     <g id="line2d_41"/>
-     <g id="text_40">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="845.812939" transform="rotate(-0 137 845.812939)">(4, 18, 10, 16, 32)</text>
-     </g>
-    </g>
-    <g id="ytick_38">
-     <g id="line2d_42"/>
-     <g id="text_41">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="864.198224" transform="rotate(-0 137 864.198224)">(4, 18, 20, 32, 16)</text>
-     </g>
-    </g>
-    <g id="ytick_39">
-     <g id="line2d_43"/>
-     <g id="text_42">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="882.583509" transform="rotate(-0 137 882.583509)">(4, 33, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_40">
-     <g id="line2d_44"/>
-     <g id="text_43">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="900.968795" transform="rotate(-0 137 900.968795)">(4, 33, 56, 96, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_41">
-     <g id="line2d_45"/>
-     <g id="text_44">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="919.35408" transform="rotate(-0 137 919.35408)">(4, 35, 20, 32, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_42">
-     <g id="line2d_46"/>
-     <g id="text_45">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="937.739365" transform="rotate(-0 137 937.739365)">(4, 49, 80, 128, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_43">
-     <g id="line2d_47"/>
-     <g id="text_46">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="956.124651" transform="rotate(-0 137 956.124651)">(4, 49, 112, 192, 4)</text>
-     </g>
-    </g>
-    <g id="ytick_44">
-     <g id="line2d_48"/>
-     <g id="text_47">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="974.509936" transform="rotate(-0 137 974.509936)">(4, 69, 40, 64, 8)</text>
-     </g>
-    </g>
-    <g id="ytick_45">
-     <g id="line2d_49"/>
-     <g id="text_48">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="992.895221" transform="rotate(-0 137 992.895221)">(4, 121, 80, 128, 4)</text>
-     </g>
-    </g>
-   </g>
-   <g id="patch_3">
-    <path d="M 144 174.272727
-L 739.752212 174.272727
-L 739.752212 186.774721
-L 144 186.774721
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_4">
-    <path d="M 144 192.658013
-L 739.752212 192.658013
-L 739.752212 205.160007
-L 144 205.160007
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_5">
-    <path d="M 144 211.043298
-L 739.752212 211.043298
-L 739.752212 223.545292
-L 144 223.545292
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_6">
-    <path d="M 144 229.428583
-L 739.752212 229.428583
-L 739.752212 241.930577
-L 144 241.930577
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_7">
-    <path d="M 144 247.813868
-L 739.752212 247.813868
-L 739.752212 260.315862
-L 144 260.315862
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_8">
-    <path d="M 144 266.199154
-L 739.752212 266.199154
-L 739.752212 278.701148
-L 144 278.701148
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_9">
-    <path d="M 144 284.584439
-L 739.752212 284.584439
-L 739.752212 297.086433
-L 144 297.086433
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_10">
-    <path d="M 144 302.969724
-L 739.752212 302.969724
-L 739.752212 315.471718
-L 144 315.471718
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_11">
-    <path d="M 144 321.355009
-L 739.752212 321.355009
-L 739.752212 333.857003
-L 144 333.857003
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_12">
-    <path d="M 144 339.740295
-L 739.752212 339.740295
-L 739.752212 352.242289
-L 144 352.242289
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_13">
-    <path d="M 144 358.12558
-L 739.752212 358.12558
-L 739.752212 370.627574
-L 144 370.627574
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_14">
-    <path d="M 144 376.510865
-L 739.752212 376.510865
-L 739.752212 389.012859
-L 144 389.012859
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_15">
-    <path d="M 144 394.89615
-L 739.752212 394.89615
-L 739.752212 407.398144
-L 144 407.398144
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_16">
-    <path d="M 144 413.281436
-L 739.752212 413.281436
-L 739.752212 425.78343
-L 144 425.78343
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_17">
-    <path d="M 144 431.666721
-L 739.752212 431.666721
-L 739.752212 444.168715
-L 144 444.168715
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_18">
-    <path d="M 144 450.052006
-L 739.752212 450.052006
-L 739.752212 462.554
-L 144 462.554
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_19">
-    <path d="M 144 468.437291
-L 739.752212 468.437291
-L 739.752212 480.939285
-L 144 480.939285
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_20">
-    <path d="M 144 486.822577
-L 739.752212 486.822577
-L 739.752212 499.324571
-L 144 499.324571
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_21">
-    <path d="M 144 505.207862
-L 739.752212 505.207862
-L 739.752212 517.709856
-L 144 517.709856
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_22">
-    <path d="M 144 523.593147
-L 739.752212 523.593147
-L 739.752212 536.095141
-L 144 536.095141
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_23">
-    <path d="M 144 541.978432
-L 739.752212 541.978432
-L 739.752212 554.480426
-L 144 554.480426
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_24">
-    <path d="M 144 560.363718
-L 739.752212 560.363718
-L 739.752212 572.865712
-L 144 572.865712
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_25">
-    <path d="M 144 578.749003
-L 739.752212 578.749003
-L 739.752212 591.250997
-L 144 591.250997
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_26">
-    <path d="M 144 597.134288
-L 739.752212 597.134288
-L 739.752212 609.636282
-L 144 609.636282
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_27">
-    <path d="M 144 615.519574
-L 739.752212 615.519574
-L 739.752212 628.021568
-L 144 628.021568
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_28">
-    <path d="M 144 633.904859
-L 739.752212 633.904859
-L 739.752212 646.406853
-L 144 646.406853
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_29">
-    <path d="M 144 652.290144
-L 739.752212 652.290144
-L 739.752212 664.792138
-L 144 664.792138
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_30">
-    <path d="M 144 670.675429
-L 739.752212 670.675429
-L 739.752212 683.177423
-L 144 683.177423
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_31">
-    <path d="M 144 689.060715
-L 739.752212 689.060715
-L 739.752212 701.562709
-L 144 701.562709
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_32">
-    <path d="M 144 707.446
-L 739.752212 707.446
-L 739.752212 719.947994
-L 144 719.947994
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_33">
-    <path d="M 144 725.831285
-L 739.752212 725.831285
-L 739.752212 738.333279
-L 144 738.333279
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_34">
-    <path d="M 144 744.21657
-L 739.752212 744.21657
-L 739.752212 756.718564
-L 144 756.718564
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_35">
-    <path d="M 144 762.601856
-L 739.752212 762.601856
-L 739.752212 775.10385
-L 144 775.10385
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_36">
-    <path d="M 144 780.987141
-L 739.752212 780.987141
-L 739.752212 793.489135
-L 144 793.489135
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_37">
-    <path d="M 144 799.372426
-L 739.752212 799.372426
-L 739.752212 811.87442
-L 144 811.87442
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_38">
-    <path d="M 144 817.757711
-L 739.752212 817.757711
-L 739.752212 830.259705
-L 144 830.259705
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_39">
-    <path d="M 144 836.142997
-L 739.752212 836.142997
-L 739.752212 848.644991
-L 144 848.644991
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_40">
-    <path d="M 144 854.528282
-L 739.752212 854.528282
-L 739.752212 867.030276
-L 144 867.030276
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_41">
-    <path d="M 144 872.913567
-L 739.752212 872.913567
-L 739.752212 885.415561
-L 144 885.415561
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_42">
-    <path d="M 144 891.298852
-L 739.752212 891.298852
-L 739.752212 903.800846
-L 144 903.800846
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_43">
-    <path d="M 144 909.684138
-L 739.752212 909.684138
-L 739.752212 922.186132
-L 144 922.186132
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_44">
-    <path d="M 144 928.069423
-L 739.752212 928.069423
-L 739.752212 940.571417
-L 144 940.571417
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_45">
-    <path d="M 144 946.454708
-L 739.752212 946.454708
-L 739.752212 958.956702
-L 144 958.956702
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_46">
-    <path d="M 144 964.839993
-L 739.752212 964.839993
-L 739.752212 977.341987
-L 144 977.341987
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_47">
-    <path d="M 144 983.225279
-L 739.752212 983.225279
-L 739.752212 995.727273
-L 144 995.727273
-z
-" clip-path="url(#p03b2f57e79)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_48">
-    <path d="M 144 1036.8
-L 144 133.2
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="patch_49">
-    <path d="M 144 1036.8
-L 817.2 1036.8
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="text_49">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="182.861615" transform="rotate(-0 743.752212 182.861615)">1</text>
-   </g>
-   <g id="text_50">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="201.2469" transform="rotate(-0 743.752212 201.2469)">1</text>
-   </g>
-   <g id="text_51">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="219.632185" transform="rotate(-0 743.752212 219.632185)">1</text>
-   </g>
-   <g id="text_52">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="238.017471" transform="rotate(-0 743.752212 238.017471)">1</text>
-   </g>
-   <g id="text_53">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="256.402756" transform="rotate(-0 743.752212 256.402756)">1</text>
-   </g>
-   <g id="text_54">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="274.788041" transform="rotate(-0 743.752212 274.788041)">1</text>
-   </g>
-   <g id="text_55">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="293.173326" transform="rotate(-0 743.752212 293.173326)">1</text>
-   </g>
-   <g id="text_56">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="311.558612" transform="rotate(-0 743.752212 311.558612)">1</text>
-   </g>
-   <g id="text_57">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="329.943897" transform="rotate(-0 743.752212 329.943897)">1</text>
-   </g>
-   <g id="text_58">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="348.329182" transform="rotate(-0 743.752212 348.329182)">1</text>
-   </g>
-   <g id="text_59">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="366.714467" transform="rotate(-0 743.752212 366.714467)">1</text>
-   </g>
-   <g id="text_60">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="385.099753" transform="rotate(-0 743.752212 385.099753)">1</text>
-   </g>
-   <g id="text_61">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="403.485038" transform="rotate(-0 743.752212 403.485038)">1</text>
-   </g>
-   <g id="text_62">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="421.870323" transform="rotate(-0 743.752212 421.870323)">1</text>
-   </g>
-   <g id="text_63">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="440.255609" transform="rotate(-0 743.752212 440.255609)">1</text>
-   </g>
-   <g id="text_64">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="458.640894" transform="rotate(-0 743.752212 458.640894)">1</text>
-   </g>
-   <g id="text_65">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="477.026179" transform="rotate(-0 743.752212 477.026179)">1</text>
-   </g>
-   <g id="text_66">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="495.411464" transform="rotate(-0 743.752212 495.411464)">1</text>
-   </g>
-   <g id="text_67">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="513.79675" transform="rotate(-0 743.752212 513.79675)">1</text>
-   </g>
-   <g id="text_68">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="532.182035" transform="rotate(-0 743.752212 532.182035)">1</text>
-   </g>
-   <g id="text_69">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="550.56732" transform="rotate(-0 743.752212 550.56732)">1</text>
-   </g>
-   <g id="text_70">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="568.952605" transform="rotate(-0 743.752212 568.952605)">1</text>
-   </g>
-   <g id="text_71">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="587.337891" transform="rotate(-0 743.752212 587.337891)">1</text>
-   </g>
-   <g id="text_72">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="605.723176" transform="rotate(-0 743.752212 605.723176)">1</text>
-   </g>
-   <g id="text_73">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="624.108461" transform="rotate(-0 743.752212 624.108461)">1</text>
-   </g>
-   <g id="text_74">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="642.493746" transform="rotate(-0 743.752212 642.493746)">1</text>
-   </g>
-   <g id="text_75">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="660.879032" transform="rotate(-0 743.752212 660.879032)">1</text>
-   </g>
-   <g id="text_76">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="679.264317" transform="rotate(-0 743.752212 679.264317)">1</text>
-   </g>
-   <g id="text_77">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="697.649602" transform="rotate(-0 743.752212 697.649602)">1</text>
-   </g>
-   <g id="text_78">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="716.034887" transform="rotate(-0 743.752212 716.034887)">1</text>
-   </g>
-   <g id="text_79">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="734.420173" transform="rotate(-0 743.752212 734.420173)">1</text>
-   </g>
-   <g id="text_80">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="752.805458" transform="rotate(-0 743.752212 752.805458)">1</text>
-   </g>
-   <g id="text_81">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="771.190743" transform="rotate(-0 743.752212 771.190743)">1</text>
-   </g>
-   <g id="text_82">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="789.576028" transform="rotate(-0 743.752212 789.576028)">1</text>
-   </g>
-   <g id="text_83">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="807.961314" transform="rotate(-0 743.752212 807.961314)">1</text>
-   </g>
-   <g id="text_84">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="826.346599" transform="rotate(-0 743.752212 826.346599)">1</text>
-   </g>
-   <g id="text_85">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="844.731884" transform="rotate(-0 743.752212 844.731884)">1</text>
-   </g>
-   <g id="text_86">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="863.11717" transform="rotate(-0 743.752212 863.11717)">1</text>
-   </g>
-   <g id="text_87">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="881.502455" transform="rotate(-0 743.752212 881.502455)">1</text>
-   </g>
-   <g id="text_88">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="899.88774" transform="rotate(-0 743.752212 899.88774)">1</text>
-   </g>
-   <g id="text_89">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="918.273025" transform="rotate(-0 743.752212 918.273025)">1</text>
-   </g>
-   <g id="text_90">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="936.658311" transform="rotate(-0 743.752212 936.658311)">1</text>
-   </g>
-   <g id="text_91">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="955.043596" transform="rotate(-0 743.752212 955.043596)">1</text>
-   </g>
-   <g id="text_92">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="973.428881" transform="rotate(-0 743.752212 973.428881)">1</text>
-   </g>
-   <g id="text_93">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="991.814166" transform="rotate(-0 743.752212 991.814166)">1</text>
-   </g>
-  </g>
-  <g id="text_94">
-   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
-  </g>
-  <g id="text_95">
-   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">neighborhood_attention_3d_d64</text>
-  </g>
-  <g id="text_96">
-   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">45 captured workloads  |  45 distinct shapes</text>
-  </g>
-  <g id="text_97">
-   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: D=64</text>
-  </g>
-  <g id="text_98">
-   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, T, Y, X, H)</text>
-  </g>
-  <g id="text_99">
-   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="1094.4" transform="rotate(-0 29.736 1094.4)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
-  </g>
- </g>
- <defs>
-  <clipPath id="p03b2f57e79">
-   <rect x="144" y="133.2" width="673.2" height="903.6"/>
-  </clipPath>
- </defs>
-</svg>
diff --git a/docs/shape-distributions/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.svg b/docs/shape-distributions/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.svg
new file mode 100644
index 0000000000000000000000000000000000000000..5301b3dee323bee33f830bb3b301d4def8f8c451
--- /dev/null
+++ b/docs/shape-distributions/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.svg
@@ -0,0 +1,258 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="375.84pt" viewBox="0 0 849.6 375.84" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: neighborhood_attention_3d_h16_d64_kt3_ky7_kx7</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: neighborhood_attention_3d_h16_d64_kt3_ky7_kx7</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 375.84
+L 849.6 375.84
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 311.04
+L 817.2 311.04
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 311.04
+L 144 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="325.637656" transform="rotate(-0 144 325.637656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 739.752212 311.04
+L 739.752212 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="325.637656" transform="rotate(-0 739.752212 325.637656)">1</text>
+     </g>
+    </g>
+    <g id="text_3">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="345.638438" transform="rotate(-0 480.6 345.638438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_5"/>
+     <g id="text_4">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="151.035384" transform="rotate(-0 137 151.035384)">(1, 9, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_6"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="169.661275" transform="rotate(-0 137 169.661275)">(1, 9, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_7"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="188.287165" transform="rotate(-0 137 188.287165)">(1, 18, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_8"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="206.913055" transform="rotate(-0 137 206.913055)">(2, 9, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_9"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="225.538945" transform="rotate(-0 137 225.538945)">(2, 9, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_10"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="244.164836" transform="rotate(-0 137 244.164836)">(2, 18, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="262.790726" transform="rotate(-0 137 262.790726)">(4, 9, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_12"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="281.416616" transform="rotate(-0 137 281.416616)">(4, 9, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_13"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="300.042506" transform="rotate(-0 137 300.042506)">(4, 18, 20, 32)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 141.283636
+L 739.752212 141.283636
+L 739.752212 153.949242
+L 144 153.949242
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 159.909527
+L 739.752212 159.909527
+L 739.752212 172.575132
+L 144 172.575132
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 178.535417
+L 739.752212 178.535417
+L 739.752212 191.201022
+L 144 191.201022
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 197.161307
+L 739.752212 197.161307
+L 739.752212 209.826912
+L 144 209.826912
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 215.787197
+L 739.752212 215.787197
+L 739.752212 228.452803
+L 144 228.452803
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 234.413088
+L 739.752212 234.413088
+L 739.752212 247.078693
+L 144 247.078693
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 253.038978
+L 739.752212 253.038978
+L 739.752212 265.704583
+L 144 265.704583
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 271.664868
+L 739.752212 271.664868
+L 739.752212 284.330473
+L 144 284.330473
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 290.290758
+L 739.752212 290.290758
+L 739.752212 302.956364
+L 144 302.956364
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 311.04
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 311.04
+L 817.2 311.04
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_13">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="149.95433" transform="rotate(-0 743.752212 149.95433)">1</text>
+   </g>
+   <g id="text_14">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="168.58022" transform="rotate(-0 743.752212 168.58022)">1</text>
+   </g>
+   <g id="text_15">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="187.20611" transform="rotate(-0 743.752212 187.20611)">1</text>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="205.832" transform="rotate(-0 743.752212 205.832)">1</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="224.457891" transform="rotate(-0 743.752212 224.457891)">1</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="243.083781" transform="rotate(-0 743.752212 243.083781)">1</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="261.709671" transform="rotate(-0 743.752212 261.709671)">1</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="280.335561" transform="rotate(-0 743.752212 280.335561)">1</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="298.961452" transform="rotate(-0 743.752212 298.961452)">1</text>
+   </g>
+  </g>
+  <g id="text_22">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_23">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">neighborhood_attention_3d_h16_d64_kt3_ky7_kx7</text>
+  </g>
+  <g id="text_24">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">9 captured workloads  |  9 distinct shapes</text>
+  </g>
+  <g id="text_25">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=16, D=64, Kt=3, Ky=7, Kx=7</text>
+  </g>
+  <g id="text_26">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, T, Y, X)</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="368.64" transform="rotate(-0 29.736 368.64)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pdd734ce0ab">
+   <rect x="144" y="133.2" width="673.2" height="177.84"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.svg b/docs/shape-distributions/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.svg
new file mode 100644
index 0000000000000000000000000000000000000000..c563b30286f8174c1109024f73e562bae25336eb
--- /dev/null
+++ b/docs/shape-distributions/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.svg
@@ -0,0 +1,258 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="375.84pt" viewBox="0 0 849.6 375.84" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: neighborhood_attention_3d_h32_d64_kt3_ky7_kx7</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: neighborhood_attention_3d_h32_d64_kt3_ky7_kx7</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 375.84
+L 849.6 375.84
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 311.04
+L 817.2 311.04
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 311.04
+L 144 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="325.637656" transform="rotate(-0 144 325.637656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 739.752212 311.04
+L 739.752212 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="325.637656" transform="rotate(-0 739.752212 325.637656)">1</text>
+     </g>
+    </g>
+    <g id="text_3">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="345.638438" transform="rotate(-0 480.6 345.638438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_5"/>
+     <g id="text_4">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="151.035384" transform="rotate(-0 137 151.035384)">(1, 9, 10, 16)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_6"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="169.661275" transform="rotate(-0 137 169.661275)">(1, 9, 14, 24)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_7"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="188.287165" transform="rotate(-0 137 188.287165)">(1, 18, 10, 16)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_8"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="206.913055" transform="rotate(-0 137 206.913055)">(2, 9, 10, 16)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_9"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="225.538945" transform="rotate(-0 137 225.538945)">(2, 9, 14, 24)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_10"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="244.164836" transform="rotate(-0 137 244.164836)">(2, 18, 10, 16)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="262.790726" transform="rotate(-0 137 262.790726)">(4, 9, 10, 16)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_12"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="281.416616" transform="rotate(-0 137 281.416616)">(4, 9, 14, 24)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_13"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="300.042506" transform="rotate(-0 137 300.042506)">(4, 18, 10, 16)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 141.283636
+L 739.752212 141.283636
+L 739.752212 153.949242
+L 144 153.949242
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 159.909527
+L 739.752212 159.909527
+L 739.752212 172.575132
+L 144 172.575132
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 178.535417
+L 739.752212 178.535417
+L 739.752212 191.201022
+L 144 191.201022
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 197.161307
+L 739.752212 197.161307
+L 739.752212 209.826912
+L 144 209.826912
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 215.787197
+L 739.752212 215.787197
+L 739.752212 228.452803
+L 144 228.452803
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 234.413088
+L 739.752212 234.413088
+L 739.752212 247.078693
+L 144 247.078693
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 253.038978
+L 739.752212 253.038978
+L 739.752212 265.704583
+L 144 265.704583
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 271.664868
+L 739.752212 271.664868
+L 739.752212 284.330473
+L 144 284.330473
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 290.290758
+L 739.752212 290.290758
+L 739.752212 302.956364
+L 144 302.956364
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 311.04
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 311.04
+L 817.2 311.04
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_13">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="149.95433" transform="rotate(-0 743.752212 149.95433)">1</text>
+   </g>
+   <g id="text_14">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="168.58022" transform="rotate(-0 743.752212 168.58022)">1</text>
+   </g>
+   <g id="text_15">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="187.20611" transform="rotate(-0 743.752212 187.20611)">1</text>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="205.832" transform="rotate(-0 743.752212 205.832)">1</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="224.457891" transform="rotate(-0 743.752212 224.457891)">1</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="243.083781" transform="rotate(-0 743.752212 243.083781)">1</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="261.709671" transform="rotate(-0 743.752212 261.709671)">1</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="280.335561" transform="rotate(-0 743.752212 280.335561)">1</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="298.961452" transform="rotate(-0 743.752212 298.961452)">1</text>
+   </g>
+  </g>
+  <g id="text_22">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_23">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">neighborhood_attention_3d_h32_d64_kt3_ky7_kx7</text>
+  </g>
+  <g id="text_24">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">9 captured workloads  |  9 distinct shapes</text>
+  </g>
+  <g id="text_25">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=32, D=64, Kt=3, Ky=7, Kx=7</text>
+  </g>
+  <g id="text_26">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, T, Y, X)</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="368.64" transform="rotate(-0 29.736 368.64)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pdd734ce0ab">
+   <rect x="144" y="133.2" width="673.2" height="177.84"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.svg b/docs/shape-distributions/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.svg
new file mode 100644
index 0000000000000000000000000000000000000000..7edf62cb59b9d027c47a7e77f4730bf96a27c654
--- /dev/null
+++ b/docs/shape-distributions/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.svg
@@ -0,0 +1,258 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="375.84pt" viewBox="0 0 849.6 375.84" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: neighborhood_attention_3d_h4_d64_kt11_ky11_kx11</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: neighborhood_attention_3d_h4_d64_kt11_ky11_kx11</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 375.84
+L 849.6 375.84
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 311.04
+L 817.2 311.04
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 311.04
+L 144 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="325.637656" transform="rotate(-0 144 325.637656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 739.752212 311.04
+L 739.752212 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="325.637656" transform="rotate(-0 739.752212 325.637656)">1</text>
+     </g>
+    </g>
+    <g id="text_3">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="345.638438" transform="rotate(-0 480.6 345.638438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_5"/>
+     <g id="text_4">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="151.035384" transform="rotate(-0 137 151.035384)">(1, 49, 80, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_6"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="169.661275" transform="rotate(-0 137 169.661275)">(1, 49, 112, 192)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_7"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="188.287165" transform="rotate(-0 137 188.287165)">(1, 121, 80, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_8"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="206.913055" transform="rotate(-0 137 206.913055)">(2, 49, 80, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_9"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="225.538945" transform="rotate(-0 137 225.538945)">(2, 49, 112, 192)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_10"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="244.164836" transform="rotate(-0 137 244.164836)">(2, 121, 80, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="262.790726" transform="rotate(-0 137 262.790726)">(4, 49, 80, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_12"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="281.416616" transform="rotate(-0 137 281.416616)">(4, 49, 112, 192)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_13"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="300.042506" transform="rotate(-0 137 300.042506)">(4, 121, 80, 128)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 141.283636
+L 739.752212 141.283636
+L 739.752212 153.949242
+L 144 153.949242
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 159.909527
+L 739.752212 159.909527
+L 739.752212 172.575132
+L 144 172.575132
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 178.535417
+L 739.752212 178.535417
+L 739.752212 191.201022
+L 144 191.201022
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 197.161307
+L 739.752212 197.161307
+L 739.752212 209.826912
+L 144 209.826912
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 215.787197
+L 739.752212 215.787197
+L 739.752212 228.452803
+L 144 228.452803
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 234.413088
+L 739.752212 234.413088
+L 739.752212 247.078693
+L 144 247.078693
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 253.038978
+L 739.752212 253.038978
+L 739.752212 265.704583
+L 144 265.704583
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 271.664868
+L 739.752212 271.664868
+L 739.752212 284.330473
+L 144 284.330473
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 290.290758
+L 739.752212 290.290758
+L 739.752212 302.956364
+L 144 302.956364
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 311.04
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 311.04
+L 817.2 311.04
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_13">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="149.95433" transform="rotate(-0 743.752212 149.95433)">1</text>
+   </g>
+   <g id="text_14">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="168.58022" transform="rotate(-0 743.752212 168.58022)">1</text>
+   </g>
+   <g id="text_15">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="187.20611" transform="rotate(-0 743.752212 187.20611)">1</text>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="205.832" transform="rotate(-0 743.752212 205.832)">1</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="224.457891" transform="rotate(-0 743.752212 224.457891)">1</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="243.083781" transform="rotate(-0 743.752212 243.083781)">1</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="261.709671" transform="rotate(-0 743.752212 261.709671)">1</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="280.335561" transform="rotate(-0 743.752212 280.335561)">1</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="298.961452" transform="rotate(-0 743.752212 298.961452)">1</text>
+   </g>
+  </g>
+  <g id="text_22">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_23">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">neighborhood_attention_3d_h4_d64_kt11_ky11_kx11</text>
+  </g>
+  <g id="text_24">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">9 captured workloads  |  9 distinct shapes</text>
+  </g>
+  <g id="text_25">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=4, D=64, Kt=11, Ky=11, Kx=11</text>
+  </g>
+  <g id="text_26">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, T, Y, X)</text>
+  </g>
+  <g id="text_27">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="368.64" transform="rotate(-0 29.736 368.64)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pdd734ce0ab">
+   <rect x="144" y="133.2" width="673.2" height="177.84"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.svg b/docs/shape-distributions/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.svg
new file mode 100644
index 0000000000000000000000000000000000000000..716b645067e2123477bfdba8c5a3c3ec7d6115a6
--- /dev/null
+++ b/docs/shape-distributions/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.svg
@@ -0,0 +1,411 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="557.28pt" viewBox="0 0 849.6 557.28" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: neighborhood_attention_3d_h8_d64_kt3_ky5_kx5</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: neighborhood_attention_3d_h8_d64_kt3_ky5_kx5</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 557.28
+L 849.6 557.28
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 492.48
+L 817.2 492.48
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 492.48
+L 144 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="507.077656" transform="rotate(-0 144 507.077656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 739.752212 492.48
+L 739.752212 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="507.077656" transform="rotate(-0 739.752212 507.077656)">1</text>
+     </g>
+    </g>
+    <g id="text_3">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="527.078438" transform="rotate(-0 480.6 527.078438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_5"/>
+     <g id="text_4">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="159.230973" transform="rotate(-0 137 159.230973)">(1, 17, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_6"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="177.704852" transform="rotate(-0 137 177.704852)">(1, 17, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_7"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="196.178731" transform="rotate(-0 137 196.178731)">(1, 33, 40, 64)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_8"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="214.65261" transform="rotate(-0 137 214.65261)">(1, 33, 56, 96)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_9"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="233.12649" transform="rotate(-0 137 233.12649)">(1, 35, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_10"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="251.600369" transform="rotate(-0 137 251.600369)">(1, 69, 40, 64)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="270.074248" transform="rotate(-0 137 270.074248)">(2, 17, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_12"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="288.548127" transform="rotate(-0 137 288.548127)">(2, 17, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_13"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="307.022006" transform="rotate(-0 137 307.022006)">(2, 33, 40, 64)</text>
+     </g>
+    </g>
+    <g id="ytick_10">
+     <g id="line2d_14"/>
+     <g id="text_13">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="325.495885" transform="rotate(-0 137 325.495885)">(2, 33, 56, 96)</text>
+     </g>
+    </g>
+    <g id="ytick_11">
+     <g id="line2d_15"/>
+     <g id="text_14">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="343.969764" transform="rotate(-0 137 343.969764)">(2, 35, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_12">
+     <g id="line2d_16"/>
+     <g id="text_15">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="362.443643" transform="rotate(-0 137 362.443643)">(2, 69, 40, 64)</text>
+     </g>
+    </g>
+    <g id="ytick_13">
+     <g id="line2d_17"/>
+     <g id="text_16">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="380.917522" transform="rotate(-0 137 380.917522)">(4, 17, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_14">
+     <g id="line2d_18"/>
+     <g id="text_17">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="399.391401" transform="rotate(-0 137 399.391401)">(4, 17, 28, 48)</text>
+     </g>
+    </g>
+    <g id="ytick_15">
+     <g id="line2d_19"/>
+     <g id="text_18">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="417.86528" transform="rotate(-0 137 417.86528)">(4, 33, 40, 64)</text>
+     </g>
+    </g>
+    <g id="ytick_16">
+     <g id="line2d_20"/>
+     <g id="text_19">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="436.339159" transform="rotate(-0 137 436.339159)">(4, 33, 56, 96)</text>
+     </g>
+    </g>
+    <g id="ytick_17">
+     <g id="line2d_21"/>
+     <g id="text_20">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="454.813038" transform="rotate(-0 137 454.813038)">(4, 35, 20, 32)</text>
+     </g>
+    </g>
+    <g id="ytick_18">
+     <g id="line2d_22"/>
+     <g id="text_21">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="473.286917" transform="rotate(-0 137 473.286917)">(4, 69, 40, 64)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 149.530909
+L 739.752212 149.530909
+L 739.752212 162.093147
+L 144 162.093147
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 168.004788
+L 739.752212 168.004788
+L 739.752212 180.567026
+L 144 180.567026
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 186.478667
+L 739.752212 186.478667
+L 739.752212 199.040905
+L 144 199.040905
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 204.952546
+L 739.752212 204.952546
+L 739.752212 217.514784
+L 144 217.514784
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 223.426425
+L 739.752212 223.426425
+L 739.752212 235.988663
+L 144 235.988663
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 241.900304
+L 739.752212 241.900304
+L 739.752212 254.462542
+L 144 254.462542
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 260.374183
+L 739.752212 260.374183
+L 739.752212 272.936421
+L 144 272.936421
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 278.848063
+L 739.752212 278.848063
+L 739.752212 291.4103
+L 144 291.4103
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 297.321942
+L 739.752212 297.321942
+L 739.752212 309.884179
+L 144 309.884179
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 315.795821
+L 739.752212 315.795821
+L 739.752212 328.358058
+L 144 328.358058
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 334.2697
+L 739.752212 334.2697
+L 739.752212 346.831937
+L 144 346.831937
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_14">
+    <path d="M 144 352.743579
+L 739.752212 352.743579
+L 739.752212 365.305817
+L 144 365.305817
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_15">
+    <path d="M 144 371.217458
+L 739.752212 371.217458
+L 739.752212 383.779696
+L 144 383.779696
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_16">
+    <path d="M 144 389.691337
+L 739.752212 389.691337
+L 739.752212 402.253575
+L 144 402.253575
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_17">
+    <path d="M 144 408.165216
+L 739.752212 408.165216
+L 739.752212 420.727454
+L 144 420.727454
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_18">
+    <path d="M 144 426.639095
+L 739.752212 426.639095
+L 739.752212 439.201333
+L 144 439.201333
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_19">
+    <path d="M 144 445.112974
+L 739.752212 445.112974
+L 739.752212 457.675212
+L 144 457.675212
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_20">
+    <path d="M 144 463.586853
+L 739.752212 463.586853
+L 739.752212 476.149091
+L 144 476.149091
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_21">
+    <path d="M 144 492.48
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_22">
+    <path d="M 144 492.48
+L 817.2 492.48
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="158.149919" transform="rotate(-0 743.752212 158.149919)">1</text>
+   </g>
+   <g id="text_23">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="176.623798" transform="rotate(-0 743.752212 176.623798)">1</text>
+   </g>
+   <g id="text_24">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="195.097677" transform="rotate(-0 743.752212 195.097677)">1</text>
+   </g>
+   <g id="text_25">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="213.571556" transform="rotate(-0 743.752212 213.571556)">1</text>
+   </g>
+   <g id="text_26">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="232.045435" transform="rotate(-0 743.752212 232.045435)">1</text>
+   </g>
+   <g id="text_27">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="250.519314" transform="rotate(-0 743.752212 250.519314)">1</text>
+   </g>
+   <g id="text_28">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="268.993193" transform="rotate(-0 743.752212 268.993193)">1</text>
+   </g>
+   <g id="text_29">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="287.467072" transform="rotate(-0 743.752212 287.467072)">1</text>
+   </g>
+   <g id="text_30">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="305.940951" transform="rotate(-0 743.752212 305.940951)">1</text>
+   </g>
+   <g id="text_31">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="324.41483" transform="rotate(-0 743.752212 324.41483)">1</text>
+   </g>
+   <g id="text_32">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="342.888709" transform="rotate(-0 743.752212 342.888709)">1</text>
+   </g>
+   <g id="text_33">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="361.362588" transform="rotate(-0 743.752212 361.362588)">1</text>
+   </g>
+   <g id="text_34">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="379.836467" transform="rotate(-0 743.752212 379.836467)">1</text>
+   </g>
+   <g id="text_35">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="398.310346" transform="rotate(-0 743.752212 398.310346)">1</text>
+   </g>
+   <g id="text_36">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="416.784225" transform="rotate(-0 743.752212 416.784225)">1</text>
+   </g>
+   <g id="text_37">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="435.258105" transform="rotate(-0 743.752212 435.258105)">1</text>
+   </g>
+   <g id="text_38">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="453.731984" transform="rotate(-0 743.752212 453.731984)">1</text>
+   </g>
+   <g id="text_39">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="472.205863" transform="rotate(-0 743.752212 472.205863)">1</text>
+   </g>
+  </g>
+  <g id="text_40">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_41">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">neighborhood_attention_3d_h8_d64_kt3_ky5_kx5</text>
+  </g>
+  <g id="text_42">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">18 captured workloads  |  18 distinct shapes</text>
+  </g>
+  <g id="text_43">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=8, D=64, Kt=3, Ky=5, Kx=5</text>
+  </g>
+  <g id="text_44">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, T, Y, X)</text>
+  </g>
+  <g id="text_45">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="550.08" transform="rotate(-0 29.736 550.08)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pa8aa21bed8">
+   <rect x="144" y="133.2" width="673.2" height="359.28"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/noncausal_dense_attention_h32.svg b/docs/shape-distributions/noncausal_dense_attention_h32.svg
deleted file mode 100644
index 4156571d4e2702af3f6e34a40cb9c59eb6ecb8d8..0000000000000000000000000000000000000000
--- a/docs/shape-distributions/noncausal_dense_attention_h32.svg
+++ /dev/null
@@ -1,976 +0,0 @@
-<?xml version="1.0" encoding="utf-8" standalone="no"?>
-<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
-  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="1162.08pt" viewBox="0 0 849.6 1162.08" xmlns="http://www.w3.org/2000/svg" version="1.1">
- <title>Shape distribution: noncausal_dense_attention_h32</title>
- <metadata>
-  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
-   <cc:Work>
-    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
-    <dc:title>Shape distribution: noncausal_dense_attention_h32</dc:title>
-    <dc:format>image/svg+xml</dc:format>
-    <dc:creator>
-     <cc:Agent>
-      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
-     </cc:Agent>
-    </dc:creator>
-   </cc:Work>
-  </rdf:RDF>
- </metadata>
- <defs>
-  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
- </defs>
- <g id="figure_1">
-  <g id="patch_1">
-   <path d="M 0 1162.08
-L 849.6 1162.08
-L 849.6 0
-L 0 0
-z
-" style="fill: #ffffff"/>
-  </g>
-  <g id="axes_1">
-   <g id="patch_2">
-    <path d="M 144 1097.28
-L 817.2 1097.28
-L 817.2 133.2
-L 144 133.2
-z
-" style="fill: #ffffff"/>
-   </g>
-   <g id="matplotlib.axis_1">
-    <g id="xtick_1">
-     <g id="line2d_1">
-      <path d="M 144 1097.28
-L 144 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_2"/>
-     <g id="text_1">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="1111.877656" transform="rotate(-0 144 1111.877656)">0</text>
-     </g>
-    </g>
-    <g id="xtick_2">
-     <g id="line2d_3">
-      <path d="M 243.292035 1097.28
-L 243.292035 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_4"/>
-     <g id="text_2">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="243.292035" y="1111.877656" transform="rotate(-0 243.292035 1111.877656)">1</text>
-     </g>
-    </g>
-    <g id="xtick_3">
-     <g id="line2d_5">
-      <path d="M 342.584071 1097.28
-L 342.584071 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_6"/>
-     <g id="text_3">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="342.584071" y="1111.877656" transform="rotate(-0 342.584071 1111.877656)">2</text>
-     </g>
-    </g>
-    <g id="xtick_4">
-     <g id="line2d_7">
-      <path d="M 441.876106 1097.28
-L 441.876106 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_8"/>
-     <g id="text_4">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="441.876106" y="1111.877656" transform="rotate(-0 441.876106 1111.877656)">3</text>
-     </g>
-    </g>
-    <g id="xtick_5">
-     <g id="line2d_9">
-      <path d="M 541.168142 1097.28
-L 541.168142 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_10"/>
-     <g id="text_5">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="541.168142" y="1111.877656" transform="rotate(-0 541.168142 1111.877656)">4</text>
-     </g>
-    </g>
-    <g id="xtick_6">
-     <g id="line2d_11">
-      <path d="M 640.460177 1097.28
-L 640.460177 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_12"/>
-     <g id="text_6">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="640.460177" y="1111.877656" transform="rotate(-0 640.460177 1111.877656)">5</text>
-     </g>
-    </g>
-    <g id="xtick_7">
-     <g id="line2d_13">
-      <path d="M 739.752212 1097.28
-L 739.752212 133.2
-" clip-path="url(#pc9dcb585f5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_14"/>
-     <g id="text_7">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="1111.877656" transform="rotate(-0 739.752212 1111.877656)">6</text>
-     </g>
-    </g>
-    <g id="text_8">
-     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="1131.878438" transform="rotate(-0 480.6 1131.878438)">Captured workloads (count)</text>
-    </g>
-   </g>
-   <g id="matplotlib.axis_2">
-    <g id="ytick_1">
-     <g id="line2d_15"/>
-     <g id="text_9">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="186.690519" transform="rotate(-0 137 186.690519)">(1, 51, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_2">
-     <g id="line2d_16"/>
-     <g id="text_10">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="205.072155" transform="rotate(-0 137 205.072155)">(1, 51, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_3">
-     <g id="line2d_17"/>
-     <g id="text_11">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="223.45379" transform="rotate(-0 137 223.45379)">(1, 51, 1120, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_4">
-     <g id="line2d_18"/>
-     <g id="text_12">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="241.835425" transform="rotate(-0 137 241.835425)">(1, 51, 2352, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_5">
-     <g id="line2d_19"/>
-     <g id="text_13">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="260.21706" transform="rotate(-0 137 260.21706)">(1, 126, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_6">
-     <g id="line2d_20"/>
-     <g id="text_14">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="278.598695" transform="rotate(-0 137 278.598695)">(1, 126, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_7">
-     <g id="line2d_21"/>
-     <g id="text_15">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="296.98033" transform="rotate(-0 137 296.98033)">(1, 126, 2560, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_8">
-     <g id="line2d_22"/>
-     <g id="text_16">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="315.361965" transform="rotate(-0 137 315.361965)">(1, 1120, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_9">
-     <g id="line2d_23"/>
-     <g id="text_17">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="333.743601" transform="rotate(-0 137 333.743601)">(1, 1120, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_10">
-     <g id="line2d_24"/>
-     <g id="text_18">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="352.125236" transform="rotate(-0 137 352.125236)">(1, 1120, 1120, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_11">
-     <g id="line2d_25"/>
-     <g id="text_19">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="370.506871" transform="rotate(-0 137 370.506871)">(1, 2352, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_12">
-     <g id="line2d_26"/>
-     <g id="text_20">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="388.888506" transform="rotate(-0 137 388.888506)">(1, 2352, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_13">
-     <g id="line2d_27"/>
-     <g id="text_21">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="407.270141" transform="rotate(-0 137 407.270141)">(1, 2352, 2352, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_14">
-     <g id="line2d_28"/>
-     <g id="text_22">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="425.651776" transform="rotate(-0 137 425.651776)">(1, 2560, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_15">
-     <g id="line2d_29"/>
-     <g id="text_23">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="444.033411" transform="rotate(-0 137 444.033411)">(1, 2560, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_16">
-     <g id="line2d_30"/>
-     <g id="text_24">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="462.415047" transform="rotate(-0 137 462.415047)">(1, 2560, 2560, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_17">
-     <g id="line2d_31"/>
-     <g id="text_25">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="480.796682" transform="rotate(-0 137 480.796682)">(2, 51, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_18">
-     <g id="line2d_32"/>
-     <g id="text_26">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="499.178317" transform="rotate(-0 137 499.178317)">(2, 51, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_19">
-     <g id="line2d_33"/>
-     <g id="text_27">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="517.559952" transform="rotate(-0 137 517.559952)">(2, 51, 1120, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_20">
-     <g id="line2d_34"/>
-     <g id="text_28">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="535.941587" transform="rotate(-0 137 535.941587)">(2, 51, 2352, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_21">
-     <g id="line2d_35"/>
-     <g id="text_29">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="554.323222" transform="rotate(-0 137 554.323222)">(2, 126, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_22">
-     <g id="line2d_36"/>
-     <g id="text_30">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="572.704857" transform="rotate(-0 137 572.704857)">(2, 126, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_23">
-     <g id="line2d_37"/>
-     <g id="text_31">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="591.086493" transform="rotate(-0 137 591.086493)">(2, 126, 2560, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_24">
-     <g id="line2d_38"/>
-     <g id="text_32">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="609.468128" transform="rotate(-0 137 609.468128)">(2, 1120, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_25">
-     <g id="line2d_39"/>
-     <g id="text_33">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="627.849763" transform="rotate(-0 137 627.849763)">(2, 1120, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_26">
-     <g id="line2d_40"/>
-     <g id="text_34">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="646.231398" transform="rotate(-0 137 646.231398)">(2, 1120, 1120, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_27">
-     <g id="line2d_41"/>
-     <g id="text_35">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="664.613033" transform="rotate(-0 137 664.613033)">(2, 2352, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_28">
-     <g id="line2d_42"/>
-     <g id="text_36">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="682.994668" transform="rotate(-0 137 682.994668)">(2, 2352, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_29">
-     <g id="line2d_43"/>
-     <g id="text_37">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="701.376303" transform="rotate(-0 137 701.376303)">(2, 2352, 2352, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_30">
-     <g id="line2d_44"/>
-     <g id="text_38">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="719.757939" transform="rotate(-0 137 719.757939)">(2, 2560, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_31">
-     <g id="line2d_45"/>
-     <g id="text_39">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="738.139574" transform="rotate(-0 137 738.139574)">(2, 2560, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_32">
-     <g id="line2d_46"/>
-     <g id="text_40">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="756.521209" transform="rotate(-0 137 756.521209)">(2, 2560, 2560, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_33">
-     <g id="line2d_47"/>
-     <g id="text_41">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="774.902844" transform="rotate(-0 137 774.902844)">(4, 51, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_34">
-     <g id="line2d_48"/>
-     <g id="text_42">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="793.284479" transform="rotate(-0 137 793.284479)">(4, 51, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_35">
-     <g id="line2d_49"/>
-     <g id="text_43">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="811.666114" transform="rotate(-0 137 811.666114)">(4, 51, 1120, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_36">
-     <g id="line2d_50"/>
-     <g id="text_44">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="830.047749" transform="rotate(-0 137 830.047749)">(4, 51, 2352, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_37">
-     <g id="line2d_51"/>
-     <g id="text_45">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="848.429385" transform="rotate(-0 137 848.429385)">(4, 126, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_38">
-     <g id="line2d_52"/>
-     <g id="text_46">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="866.81102" transform="rotate(-0 137 866.81102)">(4, 126, 1024, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_39">
-     <g id="line2d_53"/>
-     <g id="text_47">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="885.192655" transform="rotate(-0 137 885.192655)">(4, 126, 2560, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_40">
-     <g id="line2d_54"/>
-     <g id="text_48">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="903.57429" transform="rotate(-0 137 903.57429)">(4, 1120, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_41">
-     <g id="line2d_55"/>
-     <g id="text_49">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="921.955925" transform="rotate(-0 137 921.955925)">(4, 1120, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_42">
-     <g id="line2d_56"/>
-     <g id="text_50">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="940.33756" transform="rotate(-0 137 940.33756)">(4, 1120, 1120, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_43">
-     <g id="line2d_57"/>
-     <g id="text_51">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="958.719195" transform="rotate(-0 137 958.719195)">(4, 2352, 51, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_44">
-     <g id="line2d_58"/>
-     <g id="text_52">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="977.100831" transform="rotate(-0 137 977.100831)">(4, 2352, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_45">
-     <g id="line2d_59"/>
-     <g id="text_53">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="995.482466" transform="rotate(-0 137 995.482466)">(4, 2352, 2352, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_46">
-     <g id="line2d_60"/>
-     <g id="text_54">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="1013.864101" transform="rotate(-0 137 1013.864101)">(4, 2560, 126, 64)</text>
-     </g>
-    </g>
-    <g id="ytick_47">
-     <g id="line2d_61"/>
-     <g id="text_55">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="1032.245736" transform="rotate(-0 137 1032.245736)">(4, 2560, 1024, 128)</text>
-     </g>
-    </g>
-    <g id="ytick_48">
-     <g id="line2d_62"/>
-     <g id="text_56">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="1050.627371" transform="rotate(-0 137 1050.627371)">(4, 2560, 2560, 128)</text>
-     </g>
-    </g>
-   </g>
-   <g id="patch_3">
-    <path d="M 144 177.021818
-L 739.752212 177.021818
-L 739.752212 189.52133
-L 144 189.52133
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_4">
-    <path d="M 144 195.403453
-L 739.752212 195.403453
-L 739.752212 207.902965
-L 144 207.902965
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_5">
-    <path d="M 144 213.785088
-L 441.876106 213.785088
-L 441.876106 226.2846
-L 144 226.2846
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_6">
-    <path d="M 144 232.166724
-L 441.876106 232.166724
-L 441.876106 244.666236
-L 144 244.666236
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_7">
-    <path d="M 144 250.548359
-L 441.876106 250.548359
-L 441.876106 263.047871
-L 144 263.047871
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_8">
-    <path d="M 144 268.929994
-L 441.876106 268.929994
-L 441.876106 281.429506
-L 144 281.429506
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_9">
-    <path d="M 144 287.311629
-L 441.876106 287.311629
-L 441.876106 299.811141
-L 144 299.811141
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_10">
-    <path d="M 144 305.693264
-L 441.876106 305.693264
-L 441.876106 318.192776
-L 144 318.192776
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_11">
-    <path d="M 144 324.074899
-L 441.876106 324.074899
-L 441.876106 336.574411
-L 144 336.574411
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_12">
-    <path d="M 144 342.456534
-L 441.876106 342.456534
-L 441.876106 354.956046
-L 144 354.956046
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_13">
-    <path d="M 144 360.83817
-L 441.876106 360.83817
-L 441.876106 373.337682
-L 144 373.337682
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_14">
-    <path d="M 144 379.219805
-L 441.876106 379.219805
-L 441.876106 391.719317
-L 144 391.719317
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_15">
-    <path d="M 144 397.60144
-L 441.876106 397.60144
-L 441.876106 410.100952
-L 144 410.100952
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_16">
-    <path d="M 144 415.983075
-L 441.876106 415.983075
-L 441.876106 428.482587
-L 144 428.482587
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_17">
-    <path d="M 144 434.36471
-L 441.876106 434.36471
-L 441.876106 446.864222
-L 144 446.864222
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_18">
-    <path d="M 144 452.746345
-L 441.876106 452.746345
-L 441.876106 465.245857
-L 144 465.245857
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_19">
-    <path d="M 144 471.12798
-L 739.752212 471.12798
-L 739.752212 483.627492
-L 144 483.627492
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_20">
-    <path d="M 144 489.509616
-L 739.752212 489.509616
-L 739.752212 502.009128
-L 144 502.009128
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_21">
-    <path d="M 144 507.891251
-L 441.876106 507.891251
-L 441.876106 520.390763
-L 144 520.390763
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_22">
-    <path d="M 144 526.272886
-L 441.876106 526.272886
-L 441.876106 538.772398
-L 144 538.772398
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_23">
-    <path d="M 144 544.654521
-L 441.876106 544.654521
-L 441.876106 557.154033
-L 144 557.154033
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_24">
-    <path d="M 144 563.036156
-L 441.876106 563.036156
-L 441.876106 575.535668
-L 144 575.535668
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_25">
-    <path d="M 144 581.417791
-L 441.876106 581.417791
-L 441.876106 593.917303
-L 144 593.917303
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_26">
-    <path d="M 144 599.799426
-L 441.876106 599.799426
-L 441.876106 612.298938
-L 144 612.298938
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_27">
-    <path d="M 144 618.181062
-L 441.876106 618.181062
-L 441.876106 630.680574
-L 144 630.680574
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_28">
-    <path d="M 144 636.562697
-L 441.876106 636.562697
-L 441.876106 649.062209
-L 144 649.062209
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_29">
-    <path d="M 144 654.944332
-L 441.876106 654.944332
-L 441.876106 667.443844
-L 144 667.443844
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_30">
-    <path d="M 144 673.325967
-L 441.876106 673.325967
-L 441.876106 685.825479
-L 144 685.825479
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_31">
-    <path d="M 144 691.707602
-L 441.876106 691.707602
-L 441.876106 704.207114
-L 144 704.207114
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_32">
-    <path d="M 144 710.089237
-L 441.876106 710.089237
-L 441.876106 722.588749
-L 144 722.588749
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_33">
-    <path d="M 144 728.470872
-L 441.876106 728.470872
-L 441.876106 740.970384
-L 144 740.970384
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_34">
-    <path d="M 144 746.852508
-L 441.876106 746.852508
-L 441.876106 759.35202
-L 144 759.35202
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_35">
-    <path d="M 144 765.234143
-L 739.752212 765.234143
-L 739.752212 777.733655
-L 144 777.733655
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_36">
-    <path d="M 144 783.615778
-L 739.752212 783.615778
-L 739.752212 796.11529
-L 144 796.11529
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_37">
-    <path d="M 144 801.997413
-L 441.876106 801.997413
-L 441.876106 814.496925
-L 144 814.496925
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_38">
-    <path d="M 144 820.379048
-L 441.876106 820.379048
-L 441.876106 832.87856
-L 144 832.87856
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_39">
-    <path d="M 144 838.760683
-L 441.876106 838.760683
-L 441.876106 851.260195
-L 144 851.260195
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_40">
-    <path d="M 144 857.142318
-L 441.876106 857.142318
-L 441.876106 869.64183
-L 144 869.64183
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_41">
-    <path d="M 144 875.523954
-L 441.876106 875.523954
-L 441.876106 888.023466
-L 144 888.023466
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_42">
-    <path d="M 144 893.905589
-L 441.876106 893.905589
-L 441.876106 906.405101
-L 144 906.405101
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_43">
-    <path d="M 144 912.287224
-L 441.876106 912.287224
-L 441.876106 924.786736
-L 144 924.786736
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_44">
-    <path d="M 144 930.668859
-L 441.876106 930.668859
-L 441.876106 943.168371
-L 144 943.168371
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_45">
-    <path d="M 144 949.050494
-L 441.876106 949.050494
-L 441.876106 961.550006
-L 144 961.550006
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_46">
-    <path d="M 144 967.432129
-L 441.876106 967.432129
-L 441.876106 979.931641
-L 144 979.931641
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_47">
-    <path d="M 144 985.813764
-L 441.876106 985.813764
-L 441.876106 998.313276
-L 144 998.313276
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_48">
-    <path d="M 144 1004.1954
-L 441.876106 1004.1954
-L 441.876106 1016.694912
-L 144 1016.694912
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_49">
-    <path d="M 144 1022.577035
-L 441.876106 1022.577035
-L 441.876106 1035.076547
-L 144 1035.076547
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_50">
-    <path d="M 144 1040.95867
-L 441.876106 1040.95867
-L 441.876106 1053.458182
-L 144 1053.458182
-z
-" clip-path="url(#pc9dcb585f5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_51">
-    <path d="M 144 1097.28
-L 144 133.2
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="patch_52">
-    <path d="M 144 1097.28
-L 817.2 1097.28
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
-   </g>
-   <g id="text_57">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="185.609465" transform="rotate(-0 743.752212 185.609465)">6</text>
-   </g>
-   <g id="text_58">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="203.9911" transform="rotate(-0 743.752212 203.9911)">6</text>
-   </g>
-   <g id="text_59">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="222.372735" transform="rotate(-0 445.876106 222.372735)">3</text>
-   </g>
-   <g id="text_60">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="240.75437" transform="rotate(-0 445.876106 240.75437)">3</text>
-   </g>
-   <g id="text_61">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="259.136005" transform="rotate(-0 445.876106 259.136005)">3</text>
-   </g>
-   <g id="text_62">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="277.51764" transform="rotate(-0 445.876106 277.51764)">3</text>
-   </g>
-   <g id="text_63">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="295.899276" transform="rotate(-0 445.876106 295.899276)">3</text>
-   </g>
-   <g id="text_64">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="314.280911" transform="rotate(-0 445.876106 314.280911)">3</text>
-   </g>
-   <g id="text_65">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="332.662546" transform="rotate(-0 445.876106 332.662546)">3</text>
-   </g>
-   <g id="text_66">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="351.044181" transform="rotate(-0 445.876106 351.044181)">3</text>
-   </g>
-   <g id="text_67">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="369.425816" transform="rotate(-0 445.876106 369.425816)">3</text>
-   </g>
-   <g id="text_68">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="387.807451" transform="rotate(-0 445.876106 387.807451)">3</text>
-   </g>
-   <g id="text_69">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="406.189086" transform="rotate(-0 445.876106 406.189086)">3</text>
-   </g>
-   <g id="text_70">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="424.570722" transform="rotate(-0 445.876106 424.570722)">3</text>
-   </g>
-   <g id="text_71">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="442.952357" transform="rotate(-0 445.876106 442.952357)">3</text>
-   </g>
-   <g id="text_72">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="461.333992" transform="rotate(-0 445.876106 461.333992)">3</text>
-   </g>
-   <g id="text_73">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="479.715627" transform="rotate(-0 743.752212 479.715627)">6</text>
-   </g>
-   <g id="text_74">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="498.097262" transform="rotate(-0 743.752212 498.097262)">6</text>
-   </g>
-   <g id="text_75">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="516.478897" transform="rotate(-0 445.876106 516.478897)">3</text>
-   </g>
-   <g id="text_76">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="534.860532" transform="rotate(-0 445.876106 534.860532)">3</text>
-   </g>
-   <g id="text_77">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="553.242168" transform="rotate(-0 445.876106 553.242168)">3</text>
-   </g>
-   <g id="text_78">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="571.623803" transform="rotate(-0 445.876106 571.623803)">3</text>
-   </g>
-   <g id="text_79">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="590.005438" transform="rotate(-0 445.876106 590.005438)">3</text>
-   </g>
-   <g id="text_80">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="608.387073" transform="rotate(-0 445.876106 608.387073)">3</text>
-   </g>
-   <g id="text_81">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="626.768708" transform="rotate(-0 445.876106 626.768708)">3</text>
-   </g>
-   <g id="text_82">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="645.150343" transform="rotate(-0 445.876106 645.150343)">3</text>
-   </g>
-   <g id="text_83">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="663.531978" transform="rotate(-0 445.876106 663.531978)">3</text>
-   </g>
-   <g id="text_84">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="681.913614" transform="rotate(-0 445.876106 681.913614)">3</text>
-   </g>
-   <g id="text_85">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="700.295249" transform="rotate(-0 445.876106 700.295249)">3</text>
-   </g>
-   <g id="text_86">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="718.676884" transform="rotate(-0 445.876106 718.676884)">3</text>
-   </g>
-   <g id="text_87">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="737.058519" transform="rotate(-0 445.876106 737.058519)">3</text>
-   </g>
-   <g id="text_88">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="755.440154" transform="rotate(-0 445.876106 755.440154)">3</text>
-   </g>
-   <g id="text_89">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="773.821789" transform="rotate(-0 743.752212 773.821789)">6</text>
-   </g>
-   <g id="text_90">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="792.203424" transform="rotate(-0 743.752212 792.203424)">6</text>
-   </g>
-   <g id="text_91">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="810.58506" transform="rotate(-0 445.876106 810.58506)">3</text>
-   </g>
-   <g id="text_92">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="828.966695" transform="rotate(-0 445.876106 828.966695)">3</text>
-   </g>
-   <g id="text_93">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="847.34833" transform="rotate(-0 445.876106 847.34833)">3</text>
-   </g>
-   <g id="text_94">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="865.729965" transform="rotate(-0 445.876106 865.729965)">3</text>
-   </g>
-   <g id="text_95">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="884.1116" transform="rotate(-0 445.876106 884.1116)">3</text>
-   </g>
-   <g id="text_96">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="902.493235" transform="rotate(-0 445.876106 902.493235)">3</text>
-   </g>
-   <g id="text_97">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="920.87487" transform="rotate(-0 445.876106 920.87487)">3</text>
-   </g>
-   <g id="text_98">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="939.256506" transform="rotate(-0 445.876106 939.256506)">3</text>
-   </g>
-   <g id="text_99">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="957.638141" transform="rotate(-0 445.876106 957.638141)">3</text>
-   </g>
-   <g id="text_100">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="976.019776" transform="rotate(-0 445.876106 976.019776)">3</text>
-   </g>
-   <g id="text_101">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="994.401411" transform="rotate(-0 445.876106 994.401411)">3</text>
-   </g>
-   <g id="text_102">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="1012.783046" transform="rotate(-0 445.876106 1012.783046)">3</text>
-   </g>
-   <g id="text_103">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="1031.164681" transform="rotate(-0 445.876106 1031.164681)">3</text>
-   </g>
-   <g id="text_104">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="1049.546316" transform="rotate(-0 445.876106 1049.546316)">3</text>
-   </g>
-  </g>
-  <g id="text_105">
-   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
-  </g>
-  <g id="text_106">
-   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">noncausal_dense_attention_h32</text>
-  </g>
-  <g id="text_107">
-   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">162 captured workloads  |  48 distinct shapes</text>
-  </g>
-  <g id="text_108">
-   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=32</text>
-  </g>
-  <g id="text_109">
-   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, Nq, Nk, D)</text>
-  </g>
-  <g id="text_110">
-   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="1154.88" transform="rotate(-0 29.736 1154.88)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
-  </g>
- </g>
- <defs>
-  <clipPath id="pc9dcb585f5">
-   <rect x="144" y="133.2" width="673.2" height="964.08"/>
-  </clipPath>
- </defs>
-</svg>
diff --git a/docs/shape-distributions/noncausal_dense_attention_h32_d128.svg b/docs/shape-distributions/noncausal_dense_attention_h32_d128.svg
new file mode 100644
index 0000000000000000000000000000000000000000..cd1b4ff9b781b30230da158c34664f44db1c5719
--- /dev/null
+++ b/docs/shape-distributions/noncausal_dense_attention_h32_d128.svg
@@ -0,0 +1,433 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="557.28pt" viewBox="0 0 849.6 557.28" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: noncausal_dense_attention_h32_d128</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: noncausal_dense_attention_h32_d128</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 557.28
+L 849.6 557.28
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 492.48
+L 817.2 492.48
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 492.48
+L 144 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="507.077656" transform="rotate(-0 144 507.077656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 342.584071 492.48
+L 342.584071 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="342.584071" y="507.077656" transform="rotate(-0 342.584071 507.077656)">1</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_5">
+      <path d="M 541.168142 492.48
+L 541.168142 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_6"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="541.168142" y="507.077656" transform="rotate(-0 541.168142 507.077656)">2</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_7">
+      <path d="M 739.752212 492.48
+L 739.752212 133.2
+" clip-path="url(#pa8aa21bed8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_8"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="507.077656" transform="rotate(-0 739.752212 507.077656)">3</text>
+     </g>
+    </g>
+    <g id="text_5">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="527.078438" transform="rotate(-0 480.6 527.078438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_9"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="159.230973" transform="rotate(-0 137 159.230973)">(1, 1120, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_10"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="177.704852" transform="rotate(-0 137 177.704852)">(1, 1120, 1120)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_11"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="196.178731" transform="rotate(-0 137 196.178731)">(1, 2352, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_12"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="214.65261" transform="rotate(-0 137 214.65261)">(1, 2352, 2352)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_13"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="233.12649" transform="rotate(-0 137 233.12649)">(1, 2560, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_14"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="251.600369" transform="rotate(-0 137 251.600369)">(1, 2560, 2560)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_15"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="270.074248" transform="rotate(-0 137 270.074248)">(2, 1120, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_16"/>
+     <g id="text_13">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="288.548127" transform="rotate(-0 137 288.548127)">(2, 1120, 1120)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_17"/>
+     <g id="text_14">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="307.022006" transform="rotate(-0 137 307.022006)">(2, 2352, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_10">
+     <g id="line2d_18"/>
+     <g id="text_15">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="325.495885" transform="rotate(-0 137 325.495885)">(2, 2352, 2352)</text>
+     </g>
+    </g>
+    <g id="ytick_11">
+     <g id="line2d_19"/>
+     <g id="text_16">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="343.969764" transform="rotate(-0 137 343.969764)">(2, 2560, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_12">
+     <g id="line2d_20"/>
+     <g id="text_17">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="362.443643" transform="rotate(-0 137 362.443643)">(2, 2560, 2560)</text>
+     </g>
+    </g>
+    <g id="ytick_13">
+     <g id="line2d_21"/>
+     <g id="text_18">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="380.917522" transform="rotate(-0 137 380.917522)">(4, 1120, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_14">
+     <g id="line2d_22"/>
+     <g id="text_19">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="399.391401" transform="rotate(-0 137 399.391401)">(4, 1120, 1120)</text>
+     </g>
+    </g>
+    <g id="ytick_15">
+     <g id="line2d_23"/>
+     <g id="text_20">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="417.86528" transform="rotate(-0 137 417.86528)">(4, 2352, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_16">
+     <g id="line2d_24"/>
+     <g id="text_21">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="436.339159" transform="rotate(-0 137 436.339159)">(4, 2352, 2352)</text>
+     </g>
+    </g>
+    <g id="ytick_17">
+     <g id="line2d_25"/>
+     <g id="text_22">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="454.813038" transform="rotate(-0 137 454.813038)">(4, 2560, 1024)</text>
+     </g>
+    </g>
+    <g id="ytick_18">
+     <g id="line2d_26"/>
+     <g id="text_23">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="473.286917" transform="rotate(-0 137 473.286917)">(4, 2560, 2560)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 149.530909
+L 739.752212 149.530909
+L 739.752212 162.093147
+L 144 162.093147
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 168.004788
+L 739.752212 168.004788
+L 739.752212 180.567026
+L 144 180.567026
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 186.478667
+L 739.752212 186.478667
+L 739.752212 199.040905
+L 144 199.040905
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 204.952546
+L 739.752212 204.952546
+L 739.752212 217.514784
+L 144 217.514784
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 223.426425
+L 739.752212 223.426425
+L 739.752212 235.988663
+L 144 235.988663
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 241.900304
+L 739.752212 241.900304
+L 739.752212 254.462542
+L 144 254.462542
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 260.374183
+L 739.752212 260.374183
+L 739.752212 272.936421
+L 144 272.936421
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 278.848063
+L 739.752212 278.848063
+L 739.752212 291.4103
+L 144 291.4103
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 297.321942
+L 739.752212 297.321942
+L 739.752212 309.884179
+L 144 309.884179
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 315.795821
+L 739.752212 315.795821
+L 739.752212 328.358058
+L 144 328.358058
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 334.2697
+L 739.752212 334.2697
+L 739.752212 346.831937
+L 144 346.831937
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_14">
+    <path d="M 144 352.743579
+L 739.752212 352.743579
+L 739.752212 365.305817
+L 144 365.305817
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_15">
+    <path d="M 144 371.217458
+L 739.752212 371.217458
+L 739.752212 383.779696
+L 144 383.779696
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_16">
+    <path d="M 144 389.691337
+L 739.752212 389.691337
+L 739.752212 402.253575
+L 144 402.253575
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_17">
+    <path d="M 144 408.165216
+L 739.752212 408.165216
+L 739.752212 420.727454
+L 144 420.727454
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_18">
+    <path d="M 144 426.639095
+L 739.752212 426.639095
+L 739.752212 439.201333
+L 144 439.201333
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_19">
+    <path d="M 144 445.112974
+L 739.752212 445.112974
+L 739.752212 457.675212
+L 144 457.675212
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_20">
+    <path d="M 144 463.586853
+L 739.752212 463.586853
+L 739.752212 476.149091
+L 144 476.149091
+z
+" clip-path="url(#pa8aa21bed8)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_21">
+    <path d="M 144 492.48
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_22">
+    <path d="M 144 492.48
+L 817.2 492.48
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_24">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="158.149919" transform="rotate(-0 743.752212 158.149919)">3</text>
+   </g>
+   <g id="text_25">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="176.623798" transform="rotate(-0 743.752212 176.623798)">3</text>
+   </g>
+   <g id="text_26">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="195.097677" transform="rotate(-0 743.752212 195.097677)">3</text>
+   </g>
+   <g id="text_27">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="213.571556" transform="rotate(-0 743.752212 213.571556)">3</text>
+   </g>
+   <g id="text_28">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="232.045435" transform="rotate(-0 743.752212 232.045435)">3</text>
+   </g>
+   <g id="text_29">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="250.519314" transform="rotate(-0 743.752212 250.519314)">3</text>
+   </g>
+   <g id="text_30">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="268.993193" transform="rotate(-0 743.752212 268.993193)">3</text>
+   </g>
+   <g id="text_31">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="287.467072" transform="rotate(-0 743.752212 287.467072)">3</text>
+   </g>
+   <g id="text_32">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="305.940951" transform="rotate(-0 743.752212 305.940951)">3</text>
+   </g>
+   <g id="text_33">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="324.41483" transform="rotate(-0 743.752212 324.41483)">3</text>
+   </g>
+   <g id="text_34">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="342.888709" transform="rotate(-0 743.752212 342.888709)">3</text>
+   </g>
+   <g id="text_35">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="361.362588" transform="rotate(-0 743.752212 361.362588)">3</text>
+   </g>
+   <g id="text_36">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="379.836467" transform="rotate(-0 743.752212 379.836467)">3</text>
+   </g>
+   <g id="text_37">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="398.310346" transform="rotate(-0 743.752212 398.310346)">3</text>
+   </g>
+   <g id="text_38">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="416.784225" transform="rotate(-0 743.752212 416.784225)">3</text>
+   </g>
+   <g id="text_39">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="435.258105" transform="rotate(-0 743.752212 435.258105)">3</text>
+   </g>
+   <g id="text_40">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="453.731984" transform="rotate(-0 743.752212 453.731984)">3</text>
+   </g>
+   <g id="text_41">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="472.205863" transform="rotate(-0 743.752212 472.205863)">3</text>
+   </g>
+  </g>
+  <g id="text_42">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_43">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">noncausal_dense_attention_h32_d128</text>
+  </g>
+  <g id="text_44">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">54 captured workloads  |  18 distinct shapes</text>
+  </g>
+  <g id="text_45">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=32, D=128</text>
+  </g>
+  <g id="text_46">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, Nq, Nk)</text>
+  </g>
+  <g id="text_47">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="550.08" transform="rotate(-0 29.736 550.08)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pa8aa21bed8">
+   <rect x="144" y="133.2" width="673.2" height="359.28"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/sparse_mla_prefill_h64_d512.svg b/docs/shape-distributions/noncausal_dense_attention_h32_d64.svg
similarity index 52%
rename from docs/shape-distributions/sparse_mla_prefill_h64_d512.svg
rename to docs/shape-distributions/noncausal_dense_attention_h32_d64.svg
index c1805f92c509d3cd91db513af570e2cdcd590de8..17366e7d1757305cd5bd6828c2443f727ac74a4b 100644
--- a/docs/shape-distributions/sparse_mla_prefill_h64_d512.svg
+++ b/docs/shape-distributions/noncausal_dense_attention_h32_d64.svg
@@ -1,13 +1,13 @@
 <?xml version="1.0" encoding="utf-8" standalone="no"?>
 <!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
   "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="779.04pt" viewBox="0 0 849.6 779.04" xmlns="http://www.w3.org/2000/svg" version="1.1">
- <title>Shape distribution: sparse_mla_prefill_h64_d512</title>
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="799.2pt" viewBox="0 0 849.6 799.2" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: noncausal_dense_attention_h32_d64</title>
  <metadata>
   <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
    <cc:Work>
     <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
-    <dc:title>Shape distribution: sparse_mla_prefill_h64_d512</dc:title>
+    <dc:title>Shape distribution: noncausal_dense_attention_h32_d64</dc:title>
     <dc:format>image/svg+xml</dc:format>
     <dc:creator>
      <cc:Agent>
@@ -22,8 +22,8 @@
  </defs>
  <g id="figure_1">
   <g id="patch_1">
-   <path d="M 0 779.04
-L 849.6 779.04
+   <path d="M 0 799.2
+L 849.6 799.2
 L 849.6 0
 L 0 0
 z
@@ -31,8 +31,8 @@ z
   </g>
   <g id="axes_1">
    <g id="patch_2">
-    <path d="M 144 714.24
-L 817.2 714.24
+    <path d="M 144 734.4
+L 817.2 734.4
 L 817.2 133.2
 L 144 133.2
 z
@@ -41,569 +41,630 @@ z
    <g id="matplotlib.axis_1">
     <g id="xtick_1">
      <g id="line2d_1">
-      <path d="M 144 714.24
+      <path d="M 144 734.4
 L 144 133.2
-" clip-path="url(#p7160ae0ea8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
      </g>
      <g id="line2d_2"/>
      <g id="text_1">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="728.837656" transform="rotate(-0 144 728.837656)">0</text>
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="748.997656" transform="rotate(-0 144 748.997656)">0</text>
      </g>
     </g>
     <g id="xtick_2">
      <g id="line2d_3">
-      <path d="M 441.876106 714.24
-L 441.876106 133.2
-" clip-path="url(#p7160ae0ea8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+      <path d="M 243.292035 734.4
+L 243.292035 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
      </g>
      <g id="line2d_4"/>
      <g id="text_2">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="441.876106" y="728.837656" transform="rotate(-0 441.876106 728.837656)">1</text>
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="243.292035" y="748.997656" transform="rotate(-0 243.292035 748.997656)">1</text>
      </g>
     </g>
     <g id="xtick_3">
      <g id="line2d_5">
-      <path d="M 739.752212 714.24
-L 739.752212 133.2
-" clip-path="url(#p7160ae0ea8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+      <path d="M 342.584071 734.4
+L 342.584071 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
      </g>
      <g id="line2d_6"/>
      <g id="text_3">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="728.837656" transform="rotate(-0 739.752212 728.837656)">2</text>
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="342.584071" y="748.997656" transform="rotate(-0 342.584071 748.997656)">2</text>
+     </g>
+    </g>
+    <g id="xtick_4">
+     <g id="line2d_7">
+      <path d="M 441.876106 734.4
+L 441.876106 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_8"/>
+     <g id="text_4">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="441.876106" y="748.997656" transform="rotate(-0 441.876106 748.997656)">3</text>
+     </g>
+    </g>
+    <g id="xtick_5">
+     <g id="line2d_9">
+      <path d="M 541.168142 734.4
+L 541.168142 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_10"/>
+     <g id="text_5">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="541.168142" y="748.997656" transform="rotate(-0 541.168142 748.997656)">4</text>
      </g>
     </g>
-    <g id="text_4">
-     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="748.838437" transform="rotate(-0 480.6 748.838437)">Captured workloads (count)</text>
+    <g id="xtick_6">
+     <g id="line2d_11">
+      <path d="M 640.460177 734.4
+L 640.460177 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_12"/>
+     <g id="text_6">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="640.460177" y="748.997656" transform="rotate(-0 640.460177 748.997656)">5</text>
+     </g>
+    </g>
+    <g id="xtick_7">
+     <g id="line2d_13">
+      <path d="M 739.752212 734.4
+L 739.752212 133.2
+" clip-path="url(#p195062b528)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_14"/>
+     <g id="text_7">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="748.997656" transform="rotate(-0 739.752212 748.997656)">6</text>
+     </g>
+    </g>
+    <g id="text_8">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="768.998438" transform="rotate(-0 480.6 768.998438)">Captured workloads (count)</text>
     </g>
    </g>
    <g id="matplotlib.axis_2">
     <g id="ytick_1">
-     <g id="line2d_7"/>
-     <g id="text_5">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="169.291855" transform="rotate(-0 137 169.291855)">(128, 128, 128)</text>
+     <g id="line2d_15"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="170.207183" transform="rotate(-0 137 170.207183)">(1, 51, 51)</text>
      </g>
     </g>
     <g id="ytick_2">
-     <g id="line2d_8"/>
-     <g id="text_6">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="187.709504" transform="rotate(-0 137 187.709504)">(128, 192, 640)</text>
+     <g id="line2d_16"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="188.621788" transform="rotate(-0 137 188.621788)">(1, 51, 1024)</text>
      </g>
     </g>
     <g id="ytick_3">
-     <g id="line2d_9"/>
-     <g id="text_7">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="206.127154" transform="rotate(-0 137 206.127154)">(128, 256, 640)</text>
+     <g id="line2d_17"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="207.036392" transform="rotate(-0 137 207.036392)">(1, 51, 1120)</text>
      </g>
     </g>
     <g id="ytick_4">
-     <g id="line2d_10"/>
-     <g id="text_8">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="224.544803" transform="rotate(-0 137 224.544803)">(448, 575, 128)</text>
+     <g id="line2d_18"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="225.450996" transform="rotate(-0 137 225.450996)">(1, 51, 2352)</text>
      </g>
     </g>
     <g id="ytick_5">
-     <g id="line2d_11"/>
-     <g id="text_9">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="242.962452" transform="rotate(-0 137 242.962452)">(448, 71583, 640)</text>
+     <g id="line2d_19"/>
+     <g id="text_13">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="243.865601" transform="rotate(-0 137 243.865601)">(1, 126, 126)</text>
      </g>
     </g>
     <g id="ytick_6">
-     <g id="line2d_12"/>
-     <g id="text_10">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="261.380102" transform="rotate(-0 137 261.380102)">(448, 142591, 640)</text>
+     <g id="line2d_20"/>
+     <g id="text_14">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="262.280205" transform="rotate(-0 137 262.280205)">(1, 126, 1024)</text>
      </g>
     </g>
     <g id="ytick_7">
-     <g id="line2d_13"/>
-     <g id="text_11">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="279.797751" transform="rotate(-0 137 279.797751)">(1984, 2238, 128)</text>
+     <g id="line2d_21"/>
+     <g id="text_15">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="280.694809" transform="rotate(-0 137 280.694809)">(1, 126, 2560)</text>
      </g>
     </g>
     <g id="ytick_8">
-     <g id="line2d_14"/>
-     <g id="text_12">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="298.2154" transform="rotate(-0 137 298.2154)">(1984, 37662, 640)</text>
+     <g id="line2d_22"/>
+     <g id="text_16">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="299.109413" transform="rotate(-0 137 299.109413)">(1, 1120, 51)</text>
      </g>
     </g>
     <g id="ytick_9">
-     <g id="line2d_15"/>
-     <g id="text_13">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="316.63305" transform="rotate(-0 137 316.63305)">(1984, 73086, 640)</text>
+     <g id="line2d_23"/>
+     <g id="text_17">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="317.524018" transform="rotate(-0 137 317.524018)">(1, 2352, 51)</text>
      </g>
     </g>
     <g id="ytick_10">
-     <g id="line2d_16"/>
-     <g id="text_14">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="335.050699" transform="rotate(-0 137 335.050699)">(2432, 2559, 128)</text>
+     <g id="line2d_24"/>
+     <g id="text_18">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="335.938622" transform="rotate(-0 137 335.938622)">(1, 2560, 126)</text>
      </g>
     </g>
     <g id="ytick_11">
-     <g id="line2d_17"/>
-     <g id="text_15">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="353.468348" transform="rotate(-0 137 353.468348)">(2432, 497471, 640)</text>
+     <g id="line2d_25"/>
+     <g id="text_19">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="354.353226" transform="rotate(-0 137 354.353226)">(2, 51, 51)</text>
      </g>
     </g>
     <g id="ytick_12">
-     <g id="line2d_18"/>
-     <g id="text_16">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="371.885997" transform="rotate(-0 137 371.885997)">(2432, 992383, 640)</text>
+     <g id="line2d_26"/>
+     <g id="text_20">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="372.76783" transform="rotate(-0 137 372.76783)">(2, 51, 1024)</text>
      </g>
     </g>
     <g id="ytick_13">
-     <g id="line2d_19"/>
-     <g id="text_17">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="390.303647" transform="rotate(-0 137 390.303647)">(2944, 2944, 128)</text>
+     <g id="line2d_27"/>
+     <g id="text_21">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="391.182435" transform="rotate(-0 137 391.182435)">(2, 51, 1120)</text>
      </g>
     </g>
     <g id="ytick_14">
-     <g id="line2d_20"/>
-     <g id="text_18">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="408.721296" transform="rotate(-0 137 408.721296)">(2944, 4416, 640)</text>
+     <g id="line2d_28"/>
+     <g id="text_22">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="409.597039" transform="rotate(-0 137 409.597039)">(2, 51, 2352)</text>
      </g>
     </g>
     <g id="ytick_15">
-     <g id="line2d_21"/>
-     <g id="text_19">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="427.138945" transform="rotate(-0 137 427.138945)">(2944, 5888, 640)</text>
+     <g id="line2d_29"/>
+     <g id="text_23">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="428.011643" transform="rotate(-0 137 428.011643)">(2, 126, 126)</text>
      </g>
     </g>
     <g id="ytick_16">
-     <g id="line2d_22"/>
-     <g id="text_20">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="445.556595" transform="rotate(-0 137 445.556595)">(3456, 3583, 128)</text>
+     <g id="line2d_30"/>
+     <g id="text_24">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="446.426247" transform="rotate(-0 137 446.426247)">(2, 126, 1024)</text>
      </g>
     </g>
     <g id="ytick_17">
-     <g id="line2d_23"/>
-     <g id="text_21">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="463.974244" transform="rotate(-0 137 463.974244)">(3456, 278335, 640)</text>
+     <g id="line2d_31"/>
+     <g id="text_25">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="464.840852" transform="rotate(-0 137 464.840852)">(2, 126, 2560)</text>
      </g>
     </g>
     <g id="ytick_18">
-     <g id="line2d_24"/>
-     <g id="text_22">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="482.391893" transform="rotate(-0 137 482.391893)">(3456, 553087, 640)</text>
+     <g id="line2d_32"/>
+     <g id="text_26">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="483.255456" transform="rotate(-0 137 483.255456)">(2, 1120, 51)</text>
      </g>
     </g>
     <g id="ytick_19">
-     <g id="line2d_25"/>
-     <g id="text_23">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="500.809542" transform="rotate(-0 137 500.809542)">(4032, 4286, 128)</text>
+     <g id="line2d_33"/>
+     <g id="text_27">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="501.67006" transform="rotate(-0 137 501.67006)">(2, 2352, 51)</text>
      </g>
     </g>
     <g id="ytick_20">
-     <g id="line2d_26"/>
-     <g id="text_24">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="519.227192" transform="rotate(-0 137 519.227192)">(4032, 39070, 640)</text>
+     <g id="line2d_34"/>
+     <g id="text_28">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="520.084664" transform="rotate(-0 137 520.084664)">(2, 2560, 126)</text>
      </g>
     </g>
     <g id="ytick_21">
-     <g id="line2d_27"/>
-     <g id="text_25">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="537.644841" transform="rotate(-0 137 537.644841)">(4032, 73854, 640)</text>
+     <g id="line2d_35"/>
+     <g id="text_29">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="538.499269" transform="rotate(-0 137 538.499269)">(4, 51, 51)</text>
      </g>
     </g>
     <g id="ytick_22">
-     <g id="line2d_28"/>
-     <g id="text_26">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="556.06249" transform="rotate(-0 137 556.06249)">(4096, 4096, 128)</text>
+     <g id="line2d_36"/>
+     <g id="text_30">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="556.913873" transform="rotate(-0 137 556.913873)">(4, 51, 1024)</text>
      </g>
     </g>
     <g id="ytick_23">
-     <g id="line2d_29"/>
-     <g id="text_27">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="574.48014" transform="rotate(-0 137 574.48014)">(4096, 4223, 128)</text>
+     <g id="line2d_37"/>
+     <g id="text_31">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="575.328477" transform="rotate(-0 137 575.328477)">(4, 51, 1120)</text>
      </g>
     </g>
     <g id="ytick_24">
-     <g id="line2d_30"/>
-     <g id="text_28">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="592.897789" transform="rotate(-0 137 592.897789)">(4096, 6144, 640)</text>
+     <g id="line2d_38"/>
+     <g id="text_32">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="593.743082" transform="rotate(-0 137 593.743082)">(4, 51, 2352)</text>
      </g>
     </g>
     <g id="ytick_25">
-     <g id="line2d_31"/>
-     <g id="text_29">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="611.315438" transform="rotate(-0 137 611.315438)">(4096, 8192, 640)</text>
+     <g id="line2d_39"/>
+     <g id="text_33">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="612.157686" transform="rotate(-0 137 612.157686)">(4, 126, 126)</text>
      </g>
     </g>
     <g id="ytick_26">
-     <g id="line2d_32"/>
-     <g id="text_30">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="629.733088" transform="rotate(-0 137 629.733088)">(4096, 8319, 640)</text>
+     <g id="line2d_40"/>
+     <g id="text_34">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="630.57229" transform="rotate(-0 137 630.57229)">(4, 126, 1024)</text>
      </g>
     </g>
     <g id="ytick_27">
-     <g id="line2d_33"/>
-     <g id="text_31">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="648.150737" transform="rotate(-0 137 648.150737)">(4096, 12415, 640)</text>
+     <g id="line2d_41"/>
+     <g id="text_35">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="648.986894" transform="rotate(-0 137 648.986894)">(4, 126, 2560)</text>
      </g>
     </g>
     <g id="ytick_28">
-     <g id="line2d_34"/>
-     <g id="text_32">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="666.568386" transform="rotate(-0 137 666.568386)">(4096, 277247, 640)</text>
+     <g id="line2d_42"/>
+     <g id="text_36">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="667.401499" transform="rotate(-0 137 667.401499)">(4, 1120, 51)</text>
      </g>
     </g>
     <g id="ytick_29">
-     <g id="line2d_35"/>
-     <g id="text_33">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="684.986035" transform="rotate(-0 137 684.986035)">(4096, 550271, 640)</text>
+     <g id="line2d_43"/>
+     <g id="text_37">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="685.816103" transform="rotate(-0 137 685.816103)">(4, 2352, 51)</text>
+     </g>
+    </g>
+    <g id="ytick_30">
+     <g id="line2d_44"/>
+     <g id="text_38">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="704.230707" transform="rotate(-0 137 704.230707)">(4, 2560, 126)</text>
      </g>
     </g>
    </g>
    <g id="patch_3">
-    <path d="M 144 159.610909
-L 441.876106 159.610909
-L 441.876106 172.134911
-L 144 172.134911
+    <path d="M 144 160.527273
+L 739.752212 160.527273
+L 739.752212 173.049204
+L 144 173.049204
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_4">
-    <path d="M 144 178.028558
-L 441.876106 178.028558
-L 441.876106 190.55256
-L 144 190.55256
+    <path d="M 144 178.941877
+L 739.752212 178.941877
+L 739.752212 191.463808
+L 144 191.463808
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_5">
-    <path d="M 144 196.446208
-L 441.876106 196.446208
-L 441.876106 208.970209
-L 144 208.970209
+    <path d="M 144 197.356481
+L 441.876106 197.356481
+L 441.876106 209.878412
+L 144 209.878412
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_6">
-    <path d="M 144 214.863857
-L 441.876106 214.863857
-L 441.876106 227.387859
-L 144 227.387859
+    <path d="M 144 215.771086
+L 441.876106 215.771086
+L 441.876106 228.293016
+L 144 228.293016
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_7">
-    <path d="M 144 233.281506
-L 441.876106 233.281506
-L 441.876106 245.805508
-L 144 245.805508
+    <path d="M 144 234.18569
+L 441.876106 234.18569
+L 441.876106 246.707621
+L 144 246.707621
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_8">
-    <path d="M 144 251.699156
-L 441.876106 251.699156
-L 441.876106 264.223157
-L 144 264.223157
+    <path d="M 144 252.600294
+L 441.876106 252.600294
+L 441.876106 265.122225
+L 144 265.122225
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_9">
-    <path d="M 144 270.116805
-L 441.876106 270.116805
-L 441.876106 282.640806
-L 144 282.640806
+    <path d="M 144 271.014898
+L 441.876106 271.014898
+L 441.876106 283.536829
+L 144 283.536829
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_10">
-    <path d="M 144 288.534454
-L 441.876106 288.534454
-L 441.876106 301.058456
-L 144 301.058456
+    <path d="M 144 289.429503
+L 441.876106 289.429503
+L 441.876106 301.951433
+L 144 301.951433
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_11">
-    <path d="M 144 306.952103
-L 441.876106 306.952103
-L 441.876106 319.476105
-L 144 319.476105
+    <path d="M 144 307.844107
+L 441.876106 307.844107
+L 441.876106 320.366038
+L 144 320.366038
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_12">
-    <path d="M 144 325.369753
-L 441.876106 325.369753
-L 441.876106 337.893754
-L 144 337.893754
+    <path d="M 144 326.258711
+L 441.876106 326.258711
+L 441.876106 338.780642
+L 144 338.780642
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_13">
-    <path d="M 144 343.787402
-L 441.876106 343.787402
-L 441.876106 356.311404
-L 144 356.311404
+    <path d="M 144 344.673315
+L 739.752212 344.673315
+L 739.752212 357.195246
+L 144 357.195246
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_14">
-    <path d="M 144 362.205051
-L 441.876106 362.205051
-L 441.876106 374.729053
-L 144 374.729053
+    <path d="M 144 363.08792
+L 739.752212 363.08792
+L 739.752212 375.609851
+L 144 375.609851
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_15">
-    <path d="M 144 380.622701
-L 441.876106 380.622701
-L 441.876106 393.146702
-L 144 393.146702
+    <path d="M 144 381.502524
+L 441.876106 381.502524
+L 441.876106 394.024455
+L 144 394.024455
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_16">
-    <path d="M 144 399.04035
-L 441.876106 399.04035
-L 441.876106 411.564351
-L 144 411.564351
+    <path d="M 144 399.917128
+L 441.876106 399.917128
+L 441.876106 412.439059
+L 144 412.439059
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_17">
-    <path d="M 144 417.457999
-L 441.876106 417.457999
-L 441.876106 429.982001
-L 144 429.982001
+    <path d="M 144 418.331732
+L 441.876106 418.331732
+L 441.876106 430.853663
+L 144 430.853663
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_18">
-    <path d="M 144 435.875649
-L 441.876106 435.875649
-L 441.876106 448.39965
-L 144 448.39965
+    <path d="M 144 436.746337
+L 441.876106 436.746337
+L 441.876106 449.268268
+L 144 449.268268
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_19">
-    <path d="M 144 454.293298
-L 441.876106 454.293298
-L 441.876106 466.817299
-L 144 466.817299
+    <path d="M 144 455.160941
+L 441.876106 455.160941
+L 441.876106 467.682872
+L 144 467.682872
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_20">
-    <path d="M 144 472.710947
-L 441.876106 472.710947
-L 441.876106 485.234949
-L 144 485.234949
+    <path d="M 144 473.575545
+L 441.876106 473.575545
+L 441.876106 486.097476
+L 144 486.097476
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_21">
-    <path d="M 144 491.128596
-L 441.876106 491.128596
-L 441.876106 503.652598
-L 144 503.652598
+    <path d="M 144 491.990149
+L 441.876106 491.990149
+L 441.876106 504.51208
+L 144 504.51208
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_22">
-    <path d="M 144 509.546246
-L 441.876106 509.546246
-L 441.876106 522.070247
-L 144 522.070247
+    <path d="M 144 510.404754
+L 441.876106 510.404754
+L 441.876106 522.926685
+L 144 522.926685
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_23">
-    <path d="M 144 527.963895
-L 441.876106 527.963895
-L 441.876106 540.487897
-L 144 540.487897
+    <path d="M 144 528.819358
+L 739.752212 528.819358
+L 739.752212 541.341289
+L 144 541.341289
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_24">
-    <path d="M 144 546.381544
-L 441.876106 546.381544
-L 441.876106 558.905546
-L 144 558.905546
+    <path d="M 144 547.233962
+L 739.752212 547.233962
+L 739.752212 559.755893
+L 144 559.755893
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_25">
-    <path d="M 144 564.799194
-L 739.752212 564.799194
-L 739.752212 577.323195
-L 144 577.323195
+    <path d="M 144 565.648567
+L 441.876106 565.648567
+L 441.876106 578.170497
+L 144 578.170497
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_26">
-    <path d="M 144 583.216843
-L 441.876106 583.216843
-L 441.876106 595.740844
-L 144 595.740844
+    <path d="M 144 584.063171
+L 441.876106 584.063171
+L 441.876106 596.585102
+L 144 596.585102
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_27">
-    <path d="M 144 601.634492
-L 441.876106 601.634492
-L 441.876106 614.158494
-L 144 614.158494
+    <path d="M 144 602.477775
+L 441.876106 602.477775
+L 441.876106 614.999706
+L 144 614.999706
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_28">
-    <path d="M 144 620.052141
-L 441.876106 620.052141
-L 441.876106 632.576143
-L 144 632.576143
+    <path d="M 144 620.892379
+L 441.876106 620.892379
+L 441.876106 633.41431
+L 144 633.41431
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_29">
-    <path d="M 144 638.469791
-L 441.876106 638.469791
-L 441.876106 650.993792
-L 144 650.993792
+    <path d="M 144 639.306984
+L 441.876106 639.306984
+L 441.876106 651.828914
+L 144 651.828914
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_30">
-    <path d="M 144 656.88744
-L 441.876106 656.88744
-L 441.876106 669.411442
-L 144 669.411442
+    <path d="M 144 657.721588
+L 441.876106 657.721588
+L 441.876106 670.243519
+L 144 670.243519
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_31">
-    <path d="M 144 675.305089
-L 441.876106 675.305089
-L 441.876106 687.829091
-L 144 687.829091
+    <path d="M 144 676.136192
+L 441.876106 676.136192
+L 441.876106 688.658123
+L 144 688.658123
 z
-" clip-path="url(#p7160ae0ea8)" style="fill: #2563a6"/>
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_32">
-    <path d="M 144 714.24
-L 144 133.2
-" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+    <path d="M 144 694.550796
+L 441.876106 694.550796
+L 441.876106 707.072727
+L 144 707.072727
+z
+" clip-path="url(#p195062b528)" style="fill: #2563a6"/>
    </g>
    <g id="patch_33">
-    <path d="M 144 714.24
-L 817.2 714.24
+    <path d="M 144 734.4
+L 144 133.2
 " style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
    </g>
-   <g id="text_34">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="168.2108" transform="rotate(-0 445.876106 168.2108)">1</text>
-   </g>
-   <g id="text_35">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="186.62845" transform="rotate(-0 445.876106 186.62845)">1</text>
-   </g>
-   <g id="text_36">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="205.046099" transform="rotate(-0 445.876106 205.046099)">1</text>
-   </g>
-   <g id="text_37">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="223.463748" transform="rotate(-0 445.876106 223.463748)">1</text>
-   </g>
-   <g id="text_38">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="241.881398" transform="rotate(-0 445.876106 241.881398)">1</text>
+   <g id="patch_34">
+    <path d="M 144 734.4
+L 817.2 734.4
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
    </g>
    <g id="text_39">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="260.299047" transform="rotate(-0 445.876106 260.299047)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="169.126129" transform="rotate(-0 743.752212 169.126129)">6</text>
    </g>
    <g id="text_40">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="278.716696" transform="rotate(-0 445.876106 278.716696)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="187.540733" transform="rotate(-0 743.752212 187.540733)">6</text>
    </g>
    <g id="text_41">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="297.134346" transform="rotate(-0 445.876106 297.134346)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="205.955337" transform="rotate(-0 445.876106 205.955337)">3</text>
    </g>
    <g id="text_42">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="315.551995" transform="rotate(-0 445.876106 315.551995)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="224.369942" transform="rotate(-0 445.876106 224.369942)">3</text>
    </g>
    <g id="text_43">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="333.969644" transform="rotate(-0 445.876106 333.969644)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="242.784546" transform="rotate(-0 445.876106 242.784546)">3</text>
    </g>
    <g id="text_44">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="352.387293" transform="rotate(-0 445.876106 352.387293)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="261.19915" transform="rotate(-0 445.876106 261.19915)">3</text>
    </g>
    <g id="text_45">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="370.804943" transform="rotate(-0 445.876106 370.804943)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="279.613754" transform="rotate(-0 445.876106 279.613754)">3</text>
    </g>
    <g id="text_46">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="389.222592" transform="rotate(-0 445.876106 389.222592)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="298.028359" transform="rotate(-0 445.876106 298.028359)">3</text>
    </g>
    <g id="text_47">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="407.640241" transform="rotate(-0 445.876106 407.640241)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="316.442963" transform="rotate(-0 445.876106 316.442963)">3</text>
    </g>
    <g id="text_48">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="426.057891" transform="rotate(-0 445.876106 426.057891)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="334.857567" transform="rotate(-0 445.876106 334.857567)">3</text>
    </g>
    <g id="text_49">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="444.47554" transform="rotate(-0 445.876106 444.47554)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="353.272171" transform="rotate(-0 743.752212 353.272171)">6</text>
    </g>
    <g id="text_50">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="462.893189" transform="rotate(-0 445.876106 462.893189)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="371.686776" transform="rotate(-0 743.752212 371.686776)">6</text>
    </g>
    <g id="text_51">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="481.310839" transform="rotate(-0 445.876106 481.310839)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="390.10138" transform="rotate(-0 445.876106 390.10138)">3</text>
    </g>
    <g id="text_52">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="499.728488" transform="rotate(-0 445.876106 499.728488)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="408.515984" transform="rotate(-0 445.876106 408.515984)">3</text>
    </g>
    <g id="text_53">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="518.146137" transform="rotate(-0 445.876106 518.146137)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="426.930588" transform="rotate(-0 445.876106 426.930588)">3</text>
    </g>
    <g id="text_54">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="536.563786" transform="rotate(-0 445.876106 536.563786)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="445.345193" transform="rotate(-0 445.876106 445.345193)">3</text>
    </g>
    <g id="text_55">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="554.981436" transform="rotate(-0 445.876106 554.981436)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="463.759797" transform="rotate(-0 445.876106 463.759797)">3</text>
    </g>
    <g id="text_56">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="573.399085" transform="rotate(-0 743.752212 573.399085)">2</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="482.174401" transform="rotate(-0 445.876106 482.174401)">3</text>
    </g>
    <g id="text_57">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="591.816734" transform="rotate(-0 445.876106 591.816734)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="500.589006" transform="rotate(-0 445.876106 500.589006)">3</text>
    </g>
    <g id="text_58">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="610.234384" transform="rotate(-0 445.876106 610.234384)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="519.00361" transform="rotate(-0 445.876106 519.00361)">3</text>
    </g>
    <g id="text_59">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="628.652033" transform="rotate(-0 445.876106 628.652033)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="537.418214" transform="rotate(-0 743.752212 537.418214)">6</text>
    </g>
    <g id="text_60">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="647.069682" transform="rotate(-0 445.876106 647.069682)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="555.832818" transform="rotate(-0 743.752212 555.832818)">6</text>
    </g>
    <g id="text_61">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="665.487331" transform="rotate(-0 445.876106 665.487331)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="574.247423" transform="rotate(-0 445.876106 574.247423)">3</text>
    </g>
    <g id="text_62">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="683.904981" transform="rotate(-0 445.876106 683.904981)">1</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="592.662027" transform="rotate(-0 445.876106 592.662027)">3</text>
+   </g>
+   <g id="text_63">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="611.076631" transform="rotate(-0 445.876106 611.076631)">3</text>
+   </g>
+   <g id="text_64">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="629.491235" transform="rotate(-0 445.876106 629.491235)">3</text>
+   </g>
+   <g id="text_65">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="647.90584" transform="rotate(-0 445.876106 647.90584)">3</text>
+   </g>
+   <g id="text_66">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="666.320444" transform="rotate(-0 445.876106 666.320444)">3</text>
+   </g>
+   <g id="text_67">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="684.735048" transform="rotate(-0 445.876106 684.735048)">3</text>
+   </g>
+   <g id="text_68">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="703.149652" transform="rotate(-0 445.876106 703.149652)">3</text>
    </g>
   </g>
-  <g id="text_63">
+  <g id="text_69">
    <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
   </g>
-  <g id="text_64">
-   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">sparse_mla_prefill_h64_d512</text>
+  <g id="text_70">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">noncausal_dense_attention_h32_d64</text>
   </g>
-  <g id="text_65">
-   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">30 captured workloads  |  29 distinct shapes</text>
+  <g id="text_71">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">108 captured workloads  |  30 distinct shapes</text>
   </g>
-  <g id="text_66">
-   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=64, D=512, KVH=1</text>
+  <g id="text_72">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=32, D=64</text>
   </g>
-  <g id="text_67">
-   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (M, Nkv, TopK)</text>
+  <g id="text_73">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (B, Nq, Nk)</text>
   </g>
-  <g id="text_68">
-   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="771.84" transform="rotate(-0 29.736 771.84)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  <g id="text_74">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="792" transform="rotate(-0 29.736 792)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
   </g>
  </g>
  <defs>
-  <clipPath id="p7160ae0ea8">
-   <rect x="144" y="133.2" width="673.2" height="581.04"/>
+  <clipPath id="p195062b528">
+   <rect x="144" y="133.2" width="673.2" height="601.2"/>
   </clipPath>
  </defs>
 </svg>
diff --git a/docs/shape-distributions/sparse_mla_prefill_h64_d512_k128.svg b/docs/shape-distributions/sparse_mla_prefill_h64_d512_k128.svg
new file mode 100644
index 0000000000000000000000000000000000000000..902233d77b8e0dcc26b2729d871001d43e38f8bc
--- /dev/null
+++ b/docs/shape-distributions/sparse_mla_prefill_h64_d512_k128.svg
@@ -0,0 +1,269 @@
+<?xml version="1.0" encoding="utf-8" standalone="no"?>
+<!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
+  "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="375.84pt" viewBox="0 0 849.6 375.84" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: sparse_mla_prefill_h64_d512_k128</title>
+ <metadata>
+  <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
+   <cc:Work>
+    <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
+    <dc:title>Shape distribution: sparse_mla_prefill_h64_d512_k128</dc:title>
+    <dc:format>image/svg+xml</dc:format>
+    <dc:creator>
+     <cc:Agent>
+      <dc:title>Matplotlib v3.11.2, https://matplotlib.org/</dc:title>
+     </cc:Agent>
+    </dc:creator>
+   </cc:Work>
+  </rdf:RDF>
+ </metadata>
+ <defs>
+  <style type="text/css">*{stroke-linejoin: round; stroke-linecap: butt}</style>
+ </defs>
+ <g id="figure_1">
+  <g id="patch_1">
+   <path d="M 0 375.84
+L 849.6 375.84
+L 849.6 0
+L 0 0
+z
+" style="fill: #ffffff"/>
+  </g>
+  <g id="axes_1">
+   <g id="patch_2">
+    <path d="M 144 311.04
+L 817.2 311.04
+L 817.2 133.2
+L 144 133.2
+z
+" style="fill: #ffffff"/>
+   </g>
+   <g id="matplotlib.axis_1">
+    <g id="xtick_1">
+     <g id="line2d_1">
+      <path d="M 144 311.04
+L 144 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_2"/>
+     <g id="text_1">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="325.637656" transform="rotate(-0 144 325.637656)">0</text>
+     </g>
+    </g>
+    <g id="xtick_2">
+     <g id="line2d_3">
+      <path d="M 441.876106 311.04
+L 441.876106 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_4"/>
+     <g id="text_2">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="441.876106" y="325.637656" transform="rotate(-0 441.876106 325.637656)">1</text>
+     </g>
+    </g>
+    <g id="xtick_3">
+     <g id="line2d_5">
+      <path d="M 739.752212 311.04
+L 739.752212 133.2
+" clip-path="url(#pdd734ce0ab)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+     </g>
+     <g id="line2d_6"/>
+     <g id="text_3">
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="325.637656" transform="rotate(-0 739.752212 325.637656)">2</text>
+     </g>
+    </g>
+    <g id="text_4">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="345.638438" transform="rotate(-0 480.6 345.638438)">Captured workloads (count)</text>
+    </g>
+   </g>
+   <g id="matplotlib.axis_2">
+    <g id="ytick_1">
+     <g id="line2d_7"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="151.035384" transform="rotate(-0 137 151.035384)">(128, 128)</text>
+     </g>
+    </g>
+    <g id="ytick_2">
+     <g id="line2d_8"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="169.661275" transform="rotate(-0 137 169.661275)">(448, 575)</text>
+     </g>
+    </g>
+    <g id="ytick_3">
+     <g id="line2d_9"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="188.287165" transform="rotate(-0 137 188.287165)">(1984, 2238)</text>
+     </g>
+    </g>
+    <g id="ytick_4">
+     <g id="line2d_10"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="206.913055" transform="rotate(-0 137 206.913055)">(2432, 2559)</text>
+     </g>
+    </g>
+    <g id="ytick_5">
+     <g id="line2d_11"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="225.538945" transform="rotate(-0 137 225.538945)">(2944, 2944)</text>
+     </g>
+    </g>
+    <g id="ytick_6">
+     <g id="line2d_12"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="244.164836" transform="rotate(-0 137 244.164836)">(3456, 3583)</text>
+     </g>
+    </g>
+    <g id="ytick_7">
+     <g id="line2d_13"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="262.790726" transform="rotate(-0 137 262.790726)">(4032, 4286)</text>
+     </g>
+    </g>
+    <g id="ytick_8">
+     <g id="line2d_14"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="281.416616" transform="rotate(-0 137 281.416616)">(4096, 4096)</text>
+     </g>
+    </g>
+    <g id="ytick_9">
+     <g id="line2d_15"/>
+     <g id="text_13">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="300.042506" transform="rotate(-0 137 300.042506)">(4096, 4223)</text>
+     </g>
+    </g>
+   </g>
+   <g id="patch_3">
+    <path d="M 144 141.283636
+L 441.876106 141.283636
+L 441.876106 153.949242
+L 144 153.949242
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_4">
+    <path d="M 144 159.909527
+L 441.876106 159.909527
+L 441.876106 172.575132
+L 144 172.575132
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_5">
+    <path d="M 144 178.535417
+L 441.876106 178.535417
+L 441.876106 191.201022
+L 144 191.201022
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_6">
+    <path d="M 144 197.161307
+L 441.876106 197.161307
+L 441.876106 209.826912
+L 144 209.826912
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_7">
+    <path d="M 144 215.787197
+L 441.876106 215.787197
+L 441.876106 228.452803
+L 144 228.452803
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_8">
+    <path d="M 144 234.413088
+L 441.876106 234.413088
+L 441.876106 247.078693
+L 144 247.078693
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_9">
+    <path d="M 144 253.038978
+L 441.876106 253.038978
+L 441.876106 265.704583
+L 144 265.704583
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_10">
+    <path d="M 144 271.664868
+L 441.876106 271.664868
+L 441.876106 284.330473
+L 144 284.330473
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_11">
+    <path d="M 144 290.290758
+L 739.752212 290.290758
+L 739.752212 302.956364
+L 144 302.956364
+z
+" clip-path="url(#pdd734ce0ab)" style="fill: #2563a6"/>
+   </g>
+   <g id="patch_12">
+    <path d="M 144 311.04
+L 144 133.2
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="patch_13">
+    <path d="M 144 311.04
+L 817.2 311.04
+" style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
+   </g>
+   <g id="text_14">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="149.95433" transform="rotate(-0 445.876106 149.95433)">1</text>
+   </g>
+   <g id="text_15">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="168.58022" transform="rotate(-0 445.876106 168.58022)">1</text>
+   </g>
+   <g id="text_16">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="187.20611" transform="rotate(-0 445.876106 187.20611)">1</text>
+   </g>
+   <g id="text_17">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="205.832" transform="rotate(-0 445.876106 205.832)">1</text>
+   </g>
+   <g id="text_18">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="224.457891" transform="rotate(-0 445.876106 224.457891)">1</text>
+   </g>
+   <g id="text_19">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="243.083781" transform="rotate(-0 445.876106 243.083781)">1</text>
+   </g>
+   <g id="text_20">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="261.709671" transform="rotate(-0 445.876106 261.709671)">1</text>
+   </g>
+   <g id="text_21">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="445.876106" y="280.335561" transform="rotate(-0 445.876106 280.335561)">1</text>
+   </g>
+   <g id="text_22">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="298.961452" transform="rotate(-0 743.752212 298.961452)">2</text>
+   </g>
+  </g>
+  <g id="text_23">
+   <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
+  </g>
+  <g id="text_24">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">sparse_mla_prefill_h64_d512_k128</text>
+  </g>
+  <g id="text_25">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">10 captured workloads  |  9 distinct shapes</text>
+  </g>
+  <g id="text_26">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=64, D=512, KVH=1, TopK=128</text>
+  </g>
+  <g id="text_27">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (M, Nkv)</text>
+  </g>
+  <g id="text_28">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="368.64" transform="rotate(-0 29.736 368.64)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  </g>
+ </g>
+ <defs>
+  <clipPath id="pdd734ce0ab">
+   <rect x="144" y="133.2" width="673.2" height="177.84"/>
+  </clipPath>
+ </defs>
+</svg>
diff --git a/docs/shape-distributions/dense_gemm_bf16_nt.svg b/docs/shape-distributions/sparse_mla_prefill_h64_d512_k640.svg
similarity index 51%
rename from docs/shape-distributions/dense_gemm_bf16_nt.svg
rename to docs/shape-distributions/sparse_mla_prefill_h64_d512_k640.svg
index 5212f26ddb5bd88fbde8a1ea40d338aa4167ac7e..b8c77e7d51d50a759aec15a33472cb2af6d6deb6 100644
--- a/docs/shape-distributions/dense_gemm_bf16_nt.svg
+++ b/docs/shape-distributions/sparse_mla_prefill_h64_d512_k640.svg
@@ -1,13 +1,13 @@
 <?xml version="1.0" encoding="utf-8" standalone="no"?>
 <!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN"
   "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd">
-<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="617.76pt" viewBox="0 0 849.6 617.76" xmlns="http://www.w3.org/2000/svg" version="1.1">
- <title>Shape distribution: dense_gemm_bf16_nt</title>
+<svg xmlns:xlink="http://www.w3.org/1999/xlink" width="849.6pt" height="597.6pt" viewBox="0 0 849.6 597.6" xmlns="http://www.w3.org/2000/svg" version="1.1">
+ <title>Shape distribution: sparse_mla_prefill_h64_d512_k640</title>
  <metadata>
   <rdf:RDF xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:cc="http://creativecommons.org/ns#" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
    <cc:Work>
     <dc:type rdf:resource="http://purl.org/dc/dcmitype/StillImage"/>
-    <dc:title>Shape distribution: dense_gemm_bf16_nt</dc:title>
+    <dc:title>Shape distribution: sparse_mla_prefill_h64_d512_k640</dc:title>
     <dc:format>image/svg+xml</dc:format>
     <dc:creator>
      <cc:Agent>
@@ -22,8 +22,8 @@
  </defs>
  <g id="figure_1">
   <g id="patch_1">
-   <path d="M 0 617.76
-L 849.6 617.76
+   <path d="M 0 597.6
+L 849.6 597.6
 L 849.6 0
 L 0 0
 z
@@ -31,8 +31,8 @@ z
   </g>
   <g id="axes_1">
    <g id="patch_2">
-    <path d="M 144 552.96
-L 817.2 552.96
+    <path d="M 144 532.8
+L 817.2 532.8
 L 817.2 133.2
 L 144 133.2
 z
@@ -41,466 +41,405 @@ z
    <g id="matplotlib.axis_1">
     <g id="xtick_1">
      <g id="line2d_1">
-      <path d="M 144 552.96
+      <path d="M 144 532.8
 L 144 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
      </g>
      <g id="line2d_2"/>
      <g id="text_1">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="567.557656" transform="rotate(-0 144 567.557656)">0</text>
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="144" y="547.397656" transform="rotate(-0 144 547.397656)">0</text>
      </g>
     </g>
     <g id="xtick_2">
      <g id="line2d_3">
-      <path d="M 276.389381 552.96
-L 276.389381 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
+      <path d="M 739.752212 532.8
+L 739.752212 133.2
+" clip-path="url(#p40b1cb91d8)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
      </g>
      <g id="line2d_4"/>
      <g id="text_2">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="276.389381" y="567.557656" transform="rotate(-0 276.389381 567.557656)">2</text>
+      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="739.752212" y="547.397656" transform="rotate(-0 739.752212 547.397656)">1</text>
      </g>
     </g>
-    <g id="xtick_3">
-     <g id="line2d_5">
-      <path d="M 408.778761 552.96
-L 408.778761 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_6"/>
-     <g id="text_3">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="408.778761" y="567.557656" transform="rotate(-0 408.778761 567.557656)">4</text>
-     </g>
-    </g>
-    <g id="xtick_4">
-     <g id="line2d_7">
-      <path d="M 541.168142 552.96
-L 541.168142 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_8"/>
-     <g id="text_4">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="541.168142" y="567.557656" transform="rotate(-0 541.168142 567.557656)">6</text>
-     </g>
-    </g>
-    <g id="xtick_5">
-     <g id="line2d_9">
-      <path d="M 673.557522 552.96
-L 673.557522 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_10"/>
-     <g id="text_5">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="673.557522" y="567.557656" transform="rotate(-0 673.557522 567.557656)">8</text>
-     </g>
-    </g>
-    <g id="xtick_6">
-     <g id="line2d_11">
-      <path d="M 805.946903 552.96
-L 805.946903 133.2
-" clip-path="url(#p383a4d5cb5)" style="fill: none; stroke: #e3e9ef; stroke-width: 0.8; stroke-linecap: square"/>
-     </g>
-     <g id="line2d_12"/>
-     <g id="text_6">
-      <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle; fill: #334155" x="805.946903" y="567.557656" transform="rotate(-0 805.946903 567.557656)">10</text>
-     </g>
-    </g>
-    <g id="text_7">
-     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="587.558438" transform="rotate(-0 480.6 587.558438)">Captured workloads (count)</text>
+    <g id="text_3">
+     <text style="font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: middle" x="480.6" y="567.398438" transform="rotate(-0 480.6 567.398438)">Captured workloads (count)</text>
     </g>
    </g>
    <g id="matplotlib.axis_2">
     <g id="ytick_1">
-     <g id="line2d_13"/>
-     <g id="text_8">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="161.972833" transform="rotate(-0 137 161.972833)">(4096, 4096, 4096)</text>
+     <g id="line2d_5"/>
+     <g id="text_4">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="161.058635" transform="rotate(-0 137 161.058635)">(128, 192)</text>
      </g>
     </g>
     <g id="ytick_2">
-     <g id="line2d_14"/>
-     <g id="text_9">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="180.425444" transform="rotate(-0 137 180.425444)">(4096, 4096, 12288)</text>
+     <g id="line2d_6"/>
+     <g id="text_5">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="179.517615" transform="rotate(-0 137 179.517615)">(128, 256)</text>
      </g>
     </g>
     <g id="ytick_3">
-     <g id="line2d_15"/>
-     <g id="text_10">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="198.878056" transform="rotate(-0 137 198.878056)">(4096, 12288, 4096)</text>
+     <g id="line2d_7"/>
+     <g id="text_6">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="197.976595" transform="rotate(-0 137 197.976595)">(448, 71583)</text>
      </g>
     </g>
     <g id="ytick_4">
-     <g id="line2d_16"/>
-     <g id="text_11">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="217.330667" transform="rotate(-0 137 217.330667)">(4125, 4096, 4096)</text>
+     <g id="line2d_8"/>
+     <g id="text_7">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="216.435575" transform="rotate(-0 137 216.435575)">(448, 142591)</text>
      </g>
     </g>
     <g id="ytick_5">
-     <g id="line2d_17"/>
-     <g id="text_12">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="235.783278" transform="rotate(-0 137 235.783278)">(4125, 4096, 12288)</text>
+     <g id="line2d_9"/>
+     <g id="text_8">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="234.894555" transform="rotate(-0 137 234.894555)">(1984, 37662)</text>
      </g>
     </g>
     <g id="ytick_6">
-     <g id="line2d_18"/>
-     <g id="text_13">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="254.235889" transform="rotate(-0 137 254.235889)">(4125, 12288, 4096)</text>
+     <g id="line2d_10"/>
+     <g id="text_9">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="253.353535" transform="rotate(-0 137 253.353535)">(1984, 73086)</text>
      </g>
     </g>
     <g id="ytick_7">
-     <g id="line2d_19"/>
-     <g id="text_14">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="272.6885" transform="rotate(-0 137 272.6885)">(6144, 4096, 4096)</text>
+     <g id="line2d_11"/>
+     <g id="text_10">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="271.812515" transform="rotate(-0 137 271.812515)">(2432, 497471)</text>
      </g>
     </g>
     <g id="ytick_8">
-     <g id="line2d_20"/>
-     <g id="text_15">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="291.141112" transform="rotate(-0 137 291.141112)">(6144, 4096, 12288)</text>
+     <g id="line2d_12"/>
+     <g id="text_11">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="290.271495" transform="rotate(-0 137 290.271495)">(2432, 992383)</text>
      </g>
     </g>
     <g id="ytick_9">
-     <g id="line2d_21"/>
-     <g id="text_16">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="309.593723" transform="rotate(-0 137 309.593723)">(6144, 12288, 4096)</text>
+     <g id="line2d_13"/>
+     <g id="text_12">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="308.730475" transform="rotate(-0 137 308.730475)">(2944, 4416)</text>
      </g>
     </g>
     <g id="ytick_10">
-     <g id="line2d_22"/>
-     <g id="text_17">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="328.046334" transform="rotate(-0 137 328.046334)">(8192, 4096, 4096)</text>
+     <g id="line2d_14"/>
+     <g id="text_13">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="327.189455" transform="rotate(-0 137 327.189455)">(2944, 5888)</text>
      </g>
     </g>
     <g id="ytick_11">
-     <g id="line2d_23"/>
-     <g id="text_18">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="346.498945" transform="rotate(-0 137 346.498945)">(8192, 4096, 12288)</text>
+     <g id="line2d_15"/>
+     <g id="text_14">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="345.648435" transform="rotate(-0 137 345.648435)">(3456, 278335)</text>
      </g>
     </g>
     <g id="ytick_12">
-     <g id="line2d_24"/>
-     <g id="text_19">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="364.951557" transform="rotate(-0 137 364.951557)">(8192, 12288, 4096)</text>
+     <g id="line2d_16"/>
+     <g id="text_15">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="364.107415" transform="rotate(-0 137 364.107415)">(3456, 553087)</text>
      </g>
     </g>
     <g id="ytick_13">
-     <g id="line2d_25"/>
-     <g id="text_20">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="383.404168" transform="rotate(-0 137 383.404168)">(12288, 4096, 4096)</text>
+     <g id="line2d_17"/>
+     <g id="text_16">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="382.566395" transform="rotate(-0 137 382.566395)">(4032, 39070)</text>
      </g>
     </g>
     <g id="ytick_14">
-     <g id="line2d_26"/>
-     <g id="text_21">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="401.856779" transform="rotate(-0 137 401.856779)">(12288, 4096, 12288)</text>
+     <g id="line2d_18"/>
+     <g id="text_17">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="401.025375" transform="rotate(-0 137 401.025375)">(4032, 73854)</text>
      </g>
     </g>
     <g id="ytick_15">
-     <g id="line2d_27"/>
-     <g id="text_22">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="420.30939" transform="rotate(-0 137 420.30939)">(12288, 12288, 4096)</text>
+     <g id="line2d_19"/>
+     <g id="text_18">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="419.484356" transform="rotate(-0 137 419.484356)">(4096, 6144)</text>
      </g>
     </g>
     <g id="ytick_16">
-     <g id="line2d_28"/>
-     <g id="text_23">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="438.762001" transform="rotate(-0 137 438.762001)">(16384, 4096, 4096)</text>
+     <g id="line2d_20"/>
+     <g id="text_19">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="437.943336" transform="rotate(-0 137 437.943336)">(4096, 8192)</text>
      </g>
     </g>
     <g id="ytick_17">
-     <g id="line2d_29"/>
-     <g id="text_24">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="457.214613" transform="rotate(-0 137 457.214613)">(16384, 4096, 12288)</text>
+     <g id="line2d_21"/>
+     <g id="text_20">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="456.402316" transform="rotate(-0 137 456.402316)">(4096, 8319)</text>
      </g>
     </g>
     <g id="ytick_18">
-     <g id="line2d_30"/>
-     <g id="text_25">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="475.667224" transform="rotate(-0 137 475.667224)">(16384, 12288, 4096)</text>
+     <g id="line2d_22"/>
+     <g id="text_21">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="474.861296" transform="rotate(-0 137 474.861296)">(4096, 12415)</text>
      </g>
     </g>
     <g id="ytick_19">
-     <g id="line2d_31"/>
-     <g id="text_26">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="494.119835" transform="rotate(-0 137 494.119835)">(24576, 4096, 4096)</text>
+     <g id="line2d_23"/>
+     <g id="text_22">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="493.320276" transform="rotate(-0 137 493.320276)">(4096, 277247)</text>
      </g>
     </g>
     <g id="ytick_20">
-     <g id="line2d_32"/>
-     <g id="text_27">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="512.572446" transform="rotate(-0 137 512.572446)">(24576, 4096, 12288)</text>
-     </g>
-    </g>
-    <g id="ytick_21">
-     <g id="line2d_33"/>
-     <g id="text_28">
-      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="531.025057" transform="rotate(-0 137 531.025057)">(24576, 12288, 4096)</text>
+     <g id="line2d_24"/>
+     <g id="text_23">
+      <text style="font-size: 9px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: end; fill: #334155" x="137" y="511.779256" transform="rotate(-0 137 511.779256)">(4096, 550271)</text>
      </g>
     </g>
    </g>
    <g id="patch_3">
-    <path d="M 144 152.28
-L 739.752212 152.28
-L 739.752212 164.827776
-L 144 164.827776
+    <path d="M 144 151.363636
+L 739.752212 151.363636
+L 739.752212 163.915743
+L 144 163.915743
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_4">
-    <path d="M 144 170.732611
-L 739.752212 170.732611
-L 739.752212 183.280387
-L 144 183.280387
+    <path d="M 144 169.822616
+L 739.752212 169.822616
+L 739.752212 182.374723
+L 144 182.374723
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_5">
-    <path d="M 144 189.185222
-L 739.752212 189.185222
-L 739.752212 201.732998
-L 144 201.732998
+    <path d="M 144 188.281596
+L 739.752212 188.281596
+L 739.752212 200.833703
+L 144 200.833703
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_6">
-    <path d="M 144 207.637834
-L 342.584071 207.637834
-L 342.584071 220.185609
-L 144 220.185609
+    <path d="M 144 206.740576
+L 739.752212 206.740576
+L 739.752212 219.292683
+L 144 219.292683
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_7">
-    <path d="M 144 226.090445
-L 342.584071 226.090445
-L 342.584071 238.638221
-L 144 238.638221
+    <path d="M 144 225.199557
+L 739.752212 225.199557
+L 739.752212 237.751663
+L 144 237.751663
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_8">
-    <path d="M 144 244.543056
-L 342.584071 244.543056
-L 342.584071 257.090832
-L 144 257.090832
+    <path d="M 144 243.658537
+L 739.752212 243.658537
+L 739.752212 256.210643
+L 144 256.210643
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_9">
-    <path d="M 144 262.995667
-L 342.584071 262.995667
-L 342.584071 275.543443
-L 144 275.543443
+    <path d="M 144 262.117517
+L 739.752212 262.117517
+L 739.752212 274.669623
+L 144 274.669623
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_10">
-    <path d="M 144 281.448279
-L 342.584071 281.448279
-L 342.584071 293.996054
-L 144 293.996054
+    <path d="M 144 280.576497
+L 739.752212 280.576497
+L 739.752212 293.128603
+L 144 293.128603
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_11">
-    <path d="M 144 299.90089
-L 342.584071 299.90089
-L 342.584071 312.448665
-L 144 312.448665
+    <path d="M 144 299.035477
+L 739.752212 299.035477
+L 739.752212 311.587583
+L 144 311.587583
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_12">
-    <path d="M 144 318.353501
-L 541.168142 318.353501
-L 541.168142 330.901277
-L 144 330.901277
+    <path d="M 144 317.494457
+L 739.752212 317.494457
+L 739.752212 330.046563
+L 144 330.046563
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_13">
-    <path d="M 144 336.806112
-L 541.168142 336.806112
-L 541.168142 349.353888
-L 144 349.353888
+    <path d="M 144 335.953437
+L 739.752212 335.953437
+L 739.752212 348.505543
+L 144 348.505543
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_14">
-    <path d="M 144 355.258723
-L 541.168142 355.258723
-L 541.168142 367.806499
-L 144 367.806499
+    <path d="M 144 354.412417
+L 739.752212 354.412417
+L 739.752212 366.964523
+L 144 366.964523
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_15">
-    <path d="M 144 373.711335
-L 276.389381 373.711335
-L 276.389381 386.25911
-L 144 386.25911
+    <path d="M 144 372.871397
+L 739.752212 372.871397
+L 739.752212 385.423503
+L 144 385.423503
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_16">
-    <path d="M 144 392.163946
-L 276.389381 392.163946
-L 276.389381 404.711721
-L 144 404.711721
+    <path d="M 144 391.330377
+L 739.752212 391.330377
+L 739.752212 403.882483
+L 144 403.882483
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_17">
-    <path d="M 144 410.616557
-L 276.389381 410.616557
-L 276.389381 423.164333
-L 144 423.164333
+    <path d="M 144 409.789357
+L 739.752212 409.789357
+L 739.752212 422.341463
+L 144 422.341463
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_18">
-    <path d="M 144 429.069168
-L 541.168142 429.069168
-L 541.168142 441.616944
-L 144 441.616944
+    <path d="M 144 428.248337
+L 739.752212 428.248337
+L 739.752212 440.800443
+L 144 440.800443
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_19">
-    <path d="M 144 447.521779
-L 541.168142 447.521779
-L 541.168142 460.069555
-L 144 460.069555
+    <path d="M 144 446.707317
+L 739.752212 446.707317
+L 739.752212 459.259424
+L 144 459.259424
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_20">
-    <path d="M 144 465.974391
-L 541.168142 465.974391
-L 541.168142 478.522166
-L 144 478.522166
+    <path d="M 144 465.166297
+L 739.752212 465.166297
+L 739.752212 477.718404
+L 144 477.718404
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_21">
-    <path d="M 144 484.427002
-L 276.389381 484.427002
-L 276.389381 496.974778
-L 144 496.974778
+    <path d="M 144 483.625277
+L 739.752212 483.625277
+L 739.752212 496.177384
+L 144 496.177384
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_22">
-    <path d="M 144 502.879613
-L 276.389381 502.879613
-L 276.389381 515.427389
-L 144 515.427389
+    <path d="M 144 502.084257
+L 739.752212 502.084257
+L 739.752212 514.636364
+L 144 514.636364
 z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
+" clip-path="url(#p40b1cb91d8)" style="fill: #2563a6"/>
    </g>
    <g id="patch_23">
-    <path d="M 144 521.332224
-L 276.389381 521.332224
-L 276.389381 533.88
-L 144 533.88
-z
-" clip-path="url(#p383a4d5cb5)" style="fill: #2563a6"/>
-   </g>
-   <g id="patch_24">
-    <path d="M 144 552.96
+    <path d="M 144 532.8
 L 144 133.2
 " style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
    </g>
-   <g id="patch_25">
-    <path d="M 144 552.96
-L 817.2 552.96
+   <g id="patch_24">
+    <path d="M 144 532.8
+L 817.2 532.8
 " style="fill: none; stroke: #b8c5d1; stroke-width: 0.8; stroke-linejoin: miter; stroke-linecap: square"/>
    </g>
+   <g id="text_24">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="159.97758" transform="rotate(-0 743.752212 159.97758)">1</text>
+   </g>
+   <g id="text_25">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="178.43656" transform="rotate(-0 743.752212 178.43656)">1</text>
+   </g>
+   <g id="text_26">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="196.89554" transform="rotate(-0 743.752212 196.89554)">1</text>
+   </g>
+   <g id="text_27">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="215.35452" transform="rotate(-0 743.752212 215.35452)">1</text>
+   </g>
+   <g id="text_28">
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="233.8135" transform="rotate(-0 743.752212 233.8135)">1</text>
+   </g>
    <g id="text_29">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="160.891778" transform="rotate(-0 743.752212 160.891778)">9</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="252.27248" transform="rotate(-0 743.752212 252.27248)">1</text>
    </g>
    <g id="text_30">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="179.34439" transform="rotate(-0 743.752212 179.34439)">9</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="270.73146" transform="rotate(-0 743.752212 270.73146)">1</text>
    </g>
    <g id="text_31">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="197.797001" transform="rotate(-0 743.752212 197.797001)">9</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="289.190441" transform="rotate(-0 743.752212 289.190441)">1</text>
    </g>
    <g id="text_32">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="216.249612" transform="rotate(-0 346.584071 216.249612)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="307.649421" transform="rotate(-0 743.752212 307.649421)">1</text>
    </g>
    <g id="text_33">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="234.702223" transform="rotate(-0 346.584071 234.702223)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="326.108401" transform="rotate(-0 743.752212 326.108401)">1</text>
    </g>
    <g id="text_34">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="253.154835" transform="rotate(-0 346.584071 253.154835)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="344.567381" transform="rotate(-0 743.752212 344.567381)">1</text>
    </g>
    <g id="text_35">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="271.607446" transform="rotate(-0 346.584071 271.607446)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="363.026361" transform="rotate(-0 743.752212 363.026361)">1</text>
    </g>
    <g id="text_36">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="290.060057" transform="rotate(-0 346.584071 290.060057)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="381.485341" transform="rotate(-0 743.752212 381.485341)">1</text>
    </g>
    <g id="text_37">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="346.584071" y="308.512668" transform="rotate(-0 346.584071 308.512668)">3</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="399.944321" transform="rotate(-0 743.752212 399.944321)">1</text>
    </g>
    <g id="text_38">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="326.965279" transform="rotate(-0 545.168142 326.965279)">6</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="418.403301" transform="rotate(-0 743.752212 418.403301)">1</text>
    </g>
    <g id="text_39">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="345.417891" transform="rotate(-0 545.168142 345.417891)">6</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="436.862281" transform="rotate(-0 743.752212 436.862281)">1</text>
    </g>
    <g id="text_40">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="363.870502" transform="rotate(-0 545.168142 363.870502)">6</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="455.321261" transform="rotate(-0 743.752212 455.321261)">1</text>
    </g>
    <g id="text_41">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="382.323113" transform="rotate(-0 280.389381 382.323113)">2</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="473.780241" transform="rotate(-0 743.752212 473.780241)">1</text>
    </g>
    <g id="text_42">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="400.775724" transform="rotate(-0 280.389381 400.775724)">2</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="492.239221" transform="rotate(-0 743.752212 492.239221)">1</text>
    </g>
    <g id="text_43">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="419.228335" transform="rotate(-0 280.389381 419.228335)">2</text>
-   </g>
-   <g id="text_44">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="437.680947" transform="rotate(-0 545.168142 437.680947)">6</text>
-   </g>
-   <g id="text_45">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="456.133558" transform="rotate(-0 545.168142 456.133558)">6</text>
-   </g>
-   <g id="text_46">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="545.168142" y="474.586169" transform="rotate(-0 545.168142 474.586169)">6</text>
-   </g>
-   <g id="text_47">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="493.03878" transform="rotate(-0 280.389381 493.03878)">2</text>
-   </g>
-   <g id="text_48">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="511.491392" transform="rotate(-0 280.389381 511.491392)">2</text>
-   </g>
-   <g id="text_49">
-    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="280.389381" y="529.944003" transform="rotate(-0 280.389381 529.944003)">2</text>
+    <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #243c52" x="743.752212" y="510.698201" transform="rotate(-0 743.752212 510.698201)">1</text>
    </g>
   </g>
-  <g id="text_50">
+  <g id="text_44">
    <text style="font-weight: 700; font-size: 16px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #142c44" x="29.736" y="19.44" transform="rotate(-0 29.736 19.44)">Shape distribution</text>
   </g>
-  <g id="text_51">
-   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">dense_gemm_bf16_nt</text>
+  <g id="text_45">
+   <text style="font-size: 10px; font-family: 'DejaVu Sans Mono', 'Bitstream Vera Sans Mono', 'Computer Modern Typewriter', 'Andale Mono', 'Nimbus Mono L', 'Courier New', 'Courier', 'Fixed', 'Terminal', monospace; text-anchor: start; fill: #334155" x="29.736" y="37.44" transform="rotate(-0 29.736 37.44)">sparse_mla_prefill_h64_d512_k640</text>
   </g>
-  <g id="text_52">
-   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">93 captured workloads  |  21 distinct shapes</text>
+  <g id="text_46">
+   <text style="font-weight: 700; font-size: 11px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #2563a6" x="29.736" y="56.16" transform="rotate(-0 29.736 56.16)">20 captured workloads  |  20 distinct shapes</text>
   </g>
-  <g id="text_53">
-   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions:</text>
+  <g id="text_47">
+   <text style="font-size: 9px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="80.278594" transform="rotate(-0 29.736 80.278594)">Fixed dimensions: H=64, D=512, KVH=1, TopK=640</text>
   </g>
-  <g id="text_54">
-   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (M, N, K)</text>
+  <g id="text_48">
+   <text style="font-weight: 700; font-size: 10px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #334155" x="144" y="111.6" transform="rotate(-0 144 111.6)">Runtime shape: (M, Nkv)</text>
   </g>
-  <g id="text_55">
-   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="610.56" transform="rotate(-0 29.736 610.56)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
+  <g id="text_49">
+   <text style="font-size: 8.5px; font-family: 'DejaVu Sans'; text-anchor: start; fill: #526171" x="29.736" y="590.4" transform="rotate(-0 29.736 590.4)">One bin per exact runtime shape; equal spacing. Counts describe this dataset, not production traffic.</text>
   </g>
  </g>
  <defs>
-  <clipPath id="p383a4d5cb5">
-   <rect x="144" y="133.2" width="673.2" height="419.76"/>
+  <clipPath id="p40b1cb91d8">
+   <rect x="144" y="133.2" width="673.2" height="399.6"/>
   </clipPath>
  </defs>
 </svg>
diff --git a/kda/dense_gemm_bf16_nt.json b/kda/dense_gemm_bf16_nt.json
deleted file mode 100644
index 0c8febe8359a45f5221d88d6c4c8eb060fd0c6db..0000000000000000000000000000000000000000
--- a/kda/dense_gemm_bf16_nt.json
+++ /dev/null
@@ -1,30 +0,0 @@
-{
-  "definition": "definitions/gemm/dense_gemm_bf16_nt.json",
-  "workloads": [
-    "workloads/gemm/dense_gemm_bf16_nt.jsonl"
-  ],
-  "baseline": "solutions/gemm/sglang_dense_gemm_bf16_nt.json",
-  "tolerance": {
-    "atol": 0.005,
-    "rtol": 0.016,
-    "required_matched_ratio": null
-  },
-  "group_axis": "M",
-  "dev": [
-    "59e21fcd-7080-5d70-aa46-c4500984dddc",
-    "ff62c8f2-f87c-586b-b295-7d130dd1a037",
-    "1ef85b69-b938-5554-b019-1cd333110897",
-    "bc334c45-f866-5289-891d-4e12672c7e97",
-    "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505",
-    "17f61747-5e4c-5e99-a7c4-f096b147fb60",
-    "8a5941ee-6c6f-5312-a860-d0b26263e260",
-    "717bab44-772a-580e-bd31-8f449703fca5",
-    "f49b336e-43c9-5d0e-a40d-277dc443e7e1",
-    "90021c8c-9d0c-55ea-857e-e12ad4fe7979"
-  ],
-  "timing": {
-    "warmup": 5,
-    "iters": 20,
-    "trials": 2
-  }
-}
diff --git a/kda/dense_gemm_bf16_nt_n12288_k4096.json b/kda/dense_gemm_bf16_nt_n12288_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..74daa47b6e7826c57eab27062ba9e6dae8f198a3
--- /dev/null
+++ b/kda/dense_gemm_bf16_nt_n12288_k4096.json
@@ -0,0 +1,27 @@
+{
+  "definition": "definitions/gemm/dense_gemm_bf16_nt_n12288_k4096.json",
+  "workloads": [
+    "workloads/gemm/dense_gemm_bf16_nt_n12288_k4096.jsonl"
+  ],
+  "baseline": "solutions/gemm/sglang_dense_gemm_bf16_nt_n12288_k4096.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "M",
+  "dev": [
+    "ff62c8f2-f87c-586b-b295-7d130dd1a037",
+    "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505",
+    "717bab44-772a-580e-bd31-8f449703fca5",
+    "b571dcbc-df9a-515c-90ac-93ba5e03a0dd",
+    "d89ba855-bc0f-52a5-9825-2f44455164ea",
+    "6c24a9cf-df14-538e-9876-3431874e3701",
+    "8c811c7c-db79-50a4-a48d-c1e483e9418b"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 2
+  }
+}
diff --git a/kda/dense_gemm_bf16_nt_n4096_k12288.json b/kda/dense_gemm_bf16_nt_n4096_k12288.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f6c83a14580b1c43214d3811d0bdfab0d4c8217
--- /dev/null
+++ b/kda/dense_gemm_bf16_nt_n4096_k12288.json
@@ -0,0 +1,27 @@
+{
+  "definition": "definitions/gemm/dense_gemm_bf16_nt_n4096_k12288.json",
+  "workloads": [
+    "workloads/gemm/dense_gemm_bf16_nt_n4096_k12288.jsonl"
+  ],
+  "baseline": "solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k12288.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "M",
+  "dev": [
+    "1ef85b69-b938-5554-b019-1cd333110897",
+    "17f61747-5e4c-5e99-a7c4-f096b147fb60",
+    "f49b336e-43c9-5d0e-a40d-277dc443e7e1",
+    "d26eaeed-136a-5399-999a-baf0a72a8818",
+    "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a",
+    "f567a1a4-8780-53bc-bd48-6b7d561d8eb8",
+    "89b32f3e-3660-5591-8885-48dffb786922"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 2
+  }
+}
diff --git a/kda/dense_gemm_bf16_nt_n4096_k4096.json b/kda/dense_gemm_bf16_nt_n4096_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..05426a55372305c119fa298f5b47f26be0a586f0
--- /dev/null
+++ b/kda/dense_gemm_bf16_nt_n4096_k4096.json
@@ -0,0 +1,26 @@
+{
+  "definition": "definitions/gemm/dense_gemm_bf16_nt_n4096_k4096.json",
+  "workloads": [
+    "workloads/gemm/dense_gemm_bf16_nt_n4096_k4096.jsonl"
+  ],
+  "baseline": "solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k4096.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "M",
+  "dev": [
+    "59e21fcd-7080-5d70-aa46-c4500984dddc",
+    "bc334c45-f866-5289-891d-4e12672c7e97",
+    "8a5941ee-6c6f-5312-a860-d0b26263e260",
+    "90021c8c-9d0c-55ea-857e-e12ad4fe7979",
+    "4995e0dc-1c17-5f42-9530-b660dce2f09d",
+    "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 3
+  }
+}
diff --git a/kda/grouped_gemm_fp8_fp4_e384_g32.json b/kda/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
similarity index 59%
rename from kda/grouped_gemm_fp8_fp4_e384_g32.json
rename to kda/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
index ae006e9a5f2211fd0af6c88e9f022c352a03f85f..f1178b5badfcba22767d1f754554627701926211 100644
--- a/kda/grouped_gemm_fp8_fp4_e384_g32.json
+++ b/kda/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
@@ -1,26 +1,26 @@
 {
-  "definition": "definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.json",
+  "definition": "definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json",
   "workloads": [
-    "workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.jsonl"
+    "workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.jsonl"
   ],
-  "baseline": "solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_g32.json",
+  "baseline": "solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json",
   "tolerance": {
     "atol": 0.005,
     "rtol": 0.016,
     "required_matched_ratio": null
   },
-  "group_axis": "N",
+  "group_axis": "M",
   "dev": [
     "998dd71e-de7d-52ce-b64b-fd65c93b2c8d",
-    "3e554d41-bfbe-56cc-a05b-179acb8a60f4",
     "54cc9aa4-83ff-56e0-830b-a9634ca9a54c",
-    "9f6a3944-09d7-52ca-abb8-4b01960f3577",
     "d4b15562-fea6-5caf-ba58-a6ede4195c3f",
-    "22791a4a-6151-5445-806d-c3793c972208",
     "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f",
-    "59c73c1c-f9e6-54f7-b15f-f6384e355503",
     "669309b6-8328-5a7b-a2ff-dcae81dfd4f8",
-    "9972c26f-b34d-56a2-a043-a50684289467"
+    "7e7da669-03fe-5ee3-a1cf-b496e7614a56",
+    "2718713b-be58-55b3-aeb5-e3a38cce6f96",
+    "630f85b0-9ace-54b6-88a0-08bb930237f1",
+    "c24c30ba-f971-5480-9756-4906df92201e",
+    "75452890-0209-5264-bb51-e75d6c5b75c0"
   ],
   "timing": {
     "warmup": 5,
diff --git a/kda/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json b/kda/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0e40ca46082706f5797dc722047366be8a6e4ca
--- /dev/null
+++ b/kda/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
@@ -0,0 +1,30 @@
+{
+  "definition": "definitions/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json",
+  "workloads": [
+    "workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.jsonl"
+  ],
+  "baseline": "solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "M",
+  "dev": [
+    "3e554d41-bfbe-56cc-a05b-179acb8a60f4",
+    "9f6a3944-09d7-52ca-abb8-4b01960f3577",
+    "22791a4a-6151-5445-806d-c3793c972208",
+    "59c73c1c-f9e6-54f7-b15f-f6384e355503",
+    "9972c26f-b34d-56a2-a043-a50684289467",
+    "7db1bb57-9034-57c0-95b3-2a45bafb1409",
+    "852cde60-0529-5e42-83ce-0ebfe07a7c0e",
+    "acae606c-4112-548b-b870-cec7314a3f82",
+    "a51a750b-1fee-5798-ac8c-6254b2d4d4f8",
+    "b1fb4669-1248-56ff-978d-f610aa47ea2a"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 1
+  }
+}
diff --git a/kda/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json b/kda/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d5272fd60027d4b3f61b2b3492364746d22ac3d
--- /dev/null
+++ b/kda/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
@@ -0,0 +1,26 @@
+{
+  "definition": "definitions/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json",
+  "workloads": [
+    "workloads/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.jsonl"
+  ],
+  "baseline": "solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "B",
+  "dev": [
+    "ba0a26ae-0688-5564-a415-b40944dc1d2c",
+    "572aaf07-1f53-5279-b4d3-b8cc9ae232c8",
+    "5b35db05-d838-5b18-9415-11dcf35d9cff",
+    "d6c3cd98-bcbb-5014-96a6-e228b4157aab",
+    "e8f46b73-bd40-55e5-96d0-32dc0199b4f1",
+    "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 3
+  }
+}
diff --git a/kda/neighborhood_attention_3d_d64.json b/kda/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
similarity index 53%
rename from kda/neighborhood_attention_3d_d64.json
rename to kda/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
index 0a43676775e8be2e954f727bfaef5cfc09d8e639..1c3338d98effe8508893a0255ab8064ee3b2e954 100644
--- a/kda/neighborhood_attention_3d_d64.json
+++ b/kda/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
@@ -1,30 +1,26 @@
 {
-  "definition": "definitions/neighborhood_attention/neighborhood_attention_3d_d64.json",
+  "definition": "definitions/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json",
   "workloads": [
-    "workloads/neighborhood_attention/neighborhood_attention_3d_d64.jsonl"
+    "workloads/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.jsonl"
   ],
-  "baseline": "solutions/neighborhood_attention/sglang_neighborhood_attention_3d_d64.json",
+  "baseline": "solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json",
   "tolerance": {
-    "atol": 0.006,
+    "atol": 0.005,
     "rtol": 0.016,
     "required_matched_ratio": null
   },
-  "group_axis": "H",
+  "group_axis": "B",
   "dev": [
     "503c803e-a524-562c-a980-daf2bad42368",
-    "ba0a26ae-0688-5564-a415-b40944dc1d2c",
-    "95130464-5fd0-59c3-81b3-5759675a89f0",
-    "4bc37940-bacd-56bd-b86c-51279edc24de",
     "d30833f7-16da-581b-84f4-66007ed77ae3",
-    "572aaf07-1f53-5279-b4d3-b8cc9ae232c8",
-    "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687",
-    "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1",
     "d9ce70f4-0314-530a-a4d3-cb359744d16b",
-    "5b35db05-d838-5b18-9415-11dcf35d9cff"
+    "01fc8439-9178-5ae6-b4a9-8a81d737d807",
+    "867a9c5a-089e-5005-bb09-db1354c3e298",
+    "2882e24d-2dc3-5644-8e69-8c22c2b13dda"
   ],
   "timing": {
     "warmup": 5,
     "iters": 20,
-    "trials": 1
+    "trials": 3
   }
 }
diff --git a/kda/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json b/kda/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
new file mode 100644
index 0000000000000000000000000000000000000000..256d5c19696826bd42f05b64959edb7f4bcec977
--- /dev/null
+++ b/kda/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
@@ -0,0 +1,25 @@
+{
+  "definition": "definitions/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json",
+  "workloads": [
+    "workloads/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.jsonl"
+  ],
+  "baseline": "solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json",
+  "tolerance": {
+    "atol": 0.006,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "B",
+  "dev": [
+    "4bc37940-bacd-56bd-b86c-51279edc24de",
+    "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1",
+    "f0fe22fd-bde6-5d95-b486-a1233f5342de",
+    "3aacba35-3f6d-5c90-8c63-a78396ea8d44",
+    "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 1
+  }
+}
diff --git a/kda/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json b/kda/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5852776041636a679d3c710d37bae4420e2e52c
--- /dev/null
+++ b/kda/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
@@ -0,0 +1,30 @@
+{
+  "definition": "definitions/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json",
+  "workloads": [
+    "workloads/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.jsonl"
+  ],
+  "baseline": "solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "B",
+  "dev": [
+    "95130464-5fd0-59c3-81b3-5759675a89f0",
+    "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687",
+    "65bd58a7-ae16-5b36-82e1-7c801c494fc3",
+    "d25c33ca-dcd8-58c5-b326-9a5e77c4d857",
+    "fdcc196f-3464-5db0-b64f-f2e29fe569fa",
+    "b573c4cd-8e41-5e92-af6a-362e7b16f0a1",
+    "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e",
+    "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1",
+    "9866c54d-a157-5fdd-89d1-84af0fb714e6",
+    "0253ef4d-943e-5839-9cb0-8d384be40c67"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 1
+  }
+}
diff --git a/kda/noncausal_dense_attention_h32_d128.json b/kda/noncausal_dense_attention_h32_d128.json
new file mode 100644
index 0000000000000000000000000000000000000000..600befaacc4e12440176ff546850ecbc121fc401
--- /dev/null
+++ b/kda/noncausal_dense_attention_h32_d128.json
@@ -0,0 +1,30 @@
+{
+  "definition": "definitions/dense_attention/noncausal_dense_attention_h32_d128.json",
+  "workloads": [
+    "workloads/dense_attention/noncausal_dense_attention_h32_d128.jsonl"
+  ],
+  "baseline": "solutions/dense_attention/sglang_noncausal_dense_attention_h32_d128.json",
+  "tolerance": {
+    "atol": 0.005,
+    "rtol": 0.016,
+    "required_matched_ratio": 0.9999
+  },
+  "group_axis": "Nq",
+  "dev": [
+    "43bb2e12-3ee6-5e23-9192-040a8405a7ae",
+    "b76a8861-ae59-5f3b-92b4-c02019ca4941",
+    "82872221-aef0-5c8a-83c1-8c116c94a80b",
+    "a34594b2-2c8d-58cd-b443-630151fec624",
+    "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0",
+    "06131807-accf-5812-a37b-8f9ae948e0f5",
+    "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937",
+    "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf",
+    "6ca03b08-4582-5408-8c48-6f4b30225045",
+    "919246a0-b394-5215-b9c0-afda319ecefa"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 1
+  }
+}
diff --git a/kda/noncausal_dense_attention_h32.json b/kda/noncausal_dense_attention_h32_d64.json
similarity index 60%
rename from kda/noncausal_dense_attention_h32.json
rename to kda/noncausal_dense_attention_h32_d64.json
index 822256f1212a9c0bfaaac81da6ddde08f9655912..7fddf7d26348443ff91f2f938be5f0c56846513e 100644
--- a/kda/noncausal_dense_attention_h32.json
+++ b/kda/noncausal_dense_attention_h32_d64.json
@@ -1,26 +1,26 @@
 {
-  "definition": "definitions/dense_attention/noncausal_dense_attention_h32.json",
+  "definition": "definitions/dense_attention/noncausal_dense_attention_h32_d64.json",
   "workloads": [
-    "workloads/dense_attention/noncausal_dense_attention_h32.jsonl"
+    "workloads/dense_attention/noncausal_dense_attention_h32_d64.jsonl"
   ],
-  "baseline": "solutions/dense_attention/sglang_noncausal_dense_attention_h32.json",
+  "baseline": "solutions/dense_attention/sglang_noncausal_dense_attention_h32_d64.json",
   "tolerance": {
     "atol": 0.005,
     "rtol": 0.016,
     "required_matched_ratio": 0.9999
   },
-  "group_axis": "D",
+  "group_axis": "Nq",
   "dev": [
     "8867a364-4a86-5118-aeeb-e8adc8419445",
-    "43bb2e12-3ee6-5e23-9192-040a8405a7ae",
     "fe5368a5-d8b1-53b5-a6c7-24893c1a525a",
-    "b76a8861-ae59-5f3b-92b4-c02019ca4941",
     "b19d6db5-9600-536d-8f4c-543dba8b95d0",
-    "82872221-aef0-5c8a-83c1-8c116c94a80b",
     "2f9ae066-bf34-53c8-bb42-abe64c739e10",
-    "a34594b2-2c8d-58cd-b443-630151fec624",
     "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70",
-    "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0"
+    "9182253a-2949-52dd-8da2-d2ebea4c91fc",
+    "365db76e-b098-5d94-84a6-2e0791f4ea86",
+    "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0",
+    "475965cf-cbad-5baa-b685-479860697719",
+    "efa942e5-4b8c-5cb5-9554-232fa3b7b889"
   ],
   "timing": {
     "warmup": 5,
diff --git a/kda/sparse_mla_prefill_h64_d512.json b/kda/sparse_mla_prefill_h64_d512_k128.json
similarity index 60%
rename from kda/sparse_mla_prefill_h64_d512.json
rename to kda/sparse_mla_prefill_h64_d512_k128.json
index 81791993930f735cc02230a684e484a40038ac16..2388f2c5e359b9d457c88c2aecdb0bbebe74632c 100644
--- a/kda/sparse_mla_prefill_h64_d512.json
+++ b/kda/sparse_mla_prefill_h64_d512_k128.json
@@ -1,26 +1,22 @@
 {
-  "definition": "definitions/sparse_mla/sparse_mla_prefill_h64_d512.json",
+  "definition": "definitions/sparse_mla/sparse_mla_prefill_h64_d512_k128.json",
   "workloads": [
-    "workloads/sparse_mla/sparse_mla_prefill_h64_d512.jsonl"
+    "workloads/sparse_mla/sparse_mla_prefill_h64_d512_k128.jsonl"
   ],
-  "baseline": "solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512.json",
+  "baseline": "solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k128.json",
   "tolerance": {
-    "atol": 0.006,
+    "atol": 0.005,
     "rtol": 0.016,
     "required_matched_ratio": null
   },
-  "group_axis": "TopK",
+  "group_axis": "M",
   "dev": [
     "63072e3a-2bf9-5eed-bc07-c48f323b15d4",
-    "d9165058-a73b-5e77-8e61-5f92a8458b70",
     "d1602ca9-fae0-56ef-9bb2-a408fe19a22b",
-    "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9",
     "aa1f4b6e-c84d-5723-8007-090ffa76004f",
-    "495000a6-c8b0-5057-99f4-88ed4a96a4a1",
     "ec18845f-e1bb-5eba-9649-3d90d470ce9b",
-    "0b988a7a-2b52-5f49-8c99-eaa373baf797",
     "c6c6e0c9-61e2-505e-831a-fef983c9dad9",
-    "52560299-af35-5351-a4f0-cdcf97141e87"
+    "d1c89357-aa68-53ac-b13f-62d900007022"
   ],
   "timing": {
     "warmup": 5,
diff --git a/kda/sparse_mla_prefill_h64_d512_k640.json b/kda/sparse_mla_prefill_h64_d512_k640.json
new file mode 100644
index 0000000000000000000000000000000000000000..04c38942f635906eed269ec55ce4e2591505ed75
--- /dev/null
+++ b/kda/sparse_mla_prefill_h64_d512_k640.json
@@ -0,0 +1,30 @@
+{
+  "definition": "definitions/sparse_mla/sparse_mla_prefill_h64_d512_k640.json",
+  "workloads": [
+    "workloads/sparse_mla/sparse_mla_prefill_h64_d512_k640.jsonl"
+  ],
+  "baseline": "solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k640.json",
+  "tolerance": {
+    "atol": 0.006,
+    "rtol": 0.016,
+    "required_matched_ratio": null
+  },
+  "group_axis": "M",
+  "dev": [
+    "d9165058-a73b-5e77-8e61-5f92a8458b70",
+    "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9",
+    "495000a6-c8b0-5057-99f4-88ed4a96a4a1",
+    "0b988a7a-2b52-5f49-8c99-eaa373baf797",
+    "52560299-af35-5351-a4f0-cdcf97141e87",
+    "42f3b7ee-5af7-5dd1-bf58-71493057d3f5",
+    "c0332d0e-dddb-5656-a8c7-0ced515dbbfa",
+    "6385d1ff-f172-5c14-b73d-1c2b5943f045",
+    "7fd83f55-ee1c-5bec-b886-047096621576",
+    "b1ee31d7-328c-59af-a60c-189c90e2d248"
+  ],
+  "timing": {
+    "warmup": 5,
+    "iters": 20,
+    "trials": 3
+  }
+}
diff --git a/scripts/capture_ltx.py b/scripts/capture_ltx.py
index 03859835b1149f459ed963bb724f25bf4046ad2c..4ce2b807b2b043ac1f959af76156b932e5f05db1 100644
--- a/scripts/capture_ltx.py
+++ b/scripts/capture_ltx.py
@@ -75,24 +75,14 @@ def install(writer):
                     sdpa_backend=audio_sdpa if modality == "audio" else video_sdpa,
                     actual_attention_batch=int(q.shape[0]),
                     module_path="transformer_blocks.0." + _label,
-                    adapter="Preserve post-normalization/RoPE Q/K/V; expand the engine visibility mask without changing it.")
+                    adapter="Preserve post-normalization/RoPE Q/K/V; verify that the engine mask is all-visible and omit the redundant mask input.")
                 if writer.wants(name, provenance):
                     mask = kw.get("attn_mask")
-                    if mask is None:
-                        mask = torch.ones((1, 1, 1, 1), dtype=torch.bool, device=q.device)
-                    else:
-                        if mask.is_floating_point():
-                            if not bool(((mask == 0) | (mask < -1e4)).all()):
-                                raise ValueError("The boolean-mask definition cannot encode a finite attention bias")
-                            mask = mask == 0
-                        else:
-                            mask = mask.bool()
-                        if mask.ndim == 2:
-                            mask = mask[:, None, None, :]
-                        elif mask.ndim == 3:
-                            mask = mask[:, None, :, :]
-                    mask = mask.expand(q.shape[0], q.shape[2], q.shape[1], k.shape[1])
-                    writer.write(name, dict(Q=q, K=k, V=v, mask=mask, sm_scale=q.shape[-1] ** -0.5),
+                    if mask is not None:
+                        visible = mask == 0 if mask.is_floating_point() else mask.bool()
+                        if not bool(visible.all()):
+                            raise ValueError("Mask-free attention capture requires an all-visible engine mask")
+                    writer.write(name, dict(Q=q, K=k, V=v, sm_scale=q.shape[-1] ** -0.5),
                                  output.reshape_as(q), provenance)
                 return output
 
diff --git a/solutions/dense_attention/fa4_noncausal_h32.json b/solutions/dense_attention/fa4_noncausal_h32.json
deleted file mode 100644
index 6943b05f221062bacc44fccceb558294fa38172d..0000000000000000000000000000000000000000
--- a/solutions/dense_attention/fa4_noncausal_h32.json
+++ /dev/null
@@ -1,26 +0,0 @@
-{
-  "name": "fa4_noncausal_h32",
-  "definition": "noncausal_dense_attention_h32",
-  "author": "inference-kernels",
-  "description": "FlashAttention 4 CuTe DSL without a mask. For the dimension-64 audio inputs it computes in FP16 to reduce probability-rounding error and returns BF16; all conversions are included in the call. Merged by KDA from the per-shape solutions when the per-shape definitions were merged. Upstream: https://github.com/Dao-AILab/flash-attention (BSD-3-Clause).",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B200"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "torch>=2.11",
-      "flash-attn-4==4.0.0b35",
-      "nvidia-cutlass-dsl[cu13]==4.7.1",
-      "quack-kernels==0.6.5"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "import torch\nfrom flash_attn.cute.interface import flash_attn_func\n\ndef run(Q, K, V, sm_scale):\n    if Q.shape[-1] == 64:\n        output = flash_attn_func(Q.to(torch.float16), K.to(torch.float16), V.to(torch.float16), softmax_scale=sm_scale)[0]\n        return output.to(Q.dtype)\n    return flash_attn_func(Q, K, V, softmax_scale=sm_scale)[0]\n"
-    }
-  ]
-}
diff --git a/solutions/dense_attention/fa4_noncausal_h32_d128.json b/solutions/dense_attention/fa4_noncausal_h32_d128.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4d7dab360c080a39ca03f0ef6f2c4356f9d4639
--- /dev/null
+++ b/solutions/dense_attention/fa4_noncausal_h32_d128.json
@@ -0,0 +1,26 @@
+{
+  "name": "fa4_noncausal_h32_d128",
+  "definition": "noncausal_dense_attention_h32_d128",
+  "author": "inference-kernels",
+  "description": "FlashAttention 4 CuTe DSL without a mask. Upstream: https://github.com/Dao-AILab/flash-attention (BSD-3-Clause).",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B200"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "torch>=2.11",
+      "flash-attn-4==4.0.0b35",
+      "nvidia-cutlass-dsl[cu13]==4.7.1",
+      "quack-kernels==0.6.5"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "from flash_attn.cute.interface import flash_attn_func\n\ndef run(Q, K, V, sm_scale):\n    return flash_attn_func(Q, K, V, softmax_scale=sm_scale)[0]\n"
+    }
+  ]
+}
diff --git a/solutions/dense_attention/fa4_noncausal_h32_d64.json b/solutions/dense_attention/fa4_noncausal_h32_d64.json
new file mode 100644
index 0000000000000000000000000000000000000000..e29bb4d2d2cfaf6852bad0c9c8f9a85ea86f9603
--- /dev/null
+++ b/solutions/dense_attention/fa4_noncausal_h32_d64.json
@@ -0,0 +1,26 @@
+{
+  "name": "fa4_noncausal_h32_d64",
+  "definition": "noncausal_dense_attention_h32_d64",
+  "author": "inference-kernels",
+  "description": "FlashAttention 4 CuTe DSL without a mask. Converts the captured normalized audio-attention inputs to FP16 to reduce probability-rounding error, and returns BF16. All conversions are included in the call. Upstream: https://github.com/Dao-AILab/flash-attention (BSD-3-Clause).",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B200"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "torch>=2.11",
+      "flash-attn-4==4.0.0b35",
+      "nvidia-cutlass-dsl[cu13]==4.7.1",
+      "quack-kernels==0.6.5"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "import torch\nfrom flash_attn.cute.interface import flash_attn_func\n\ndef run(Q, K, V, sm_scale):\n    output = flash_attn_func(Q.to(torch.float16), K.to(torch.float16), V.to(torch.float16), softmax_scale=sm_scale)[0]\n    return output.to(Q.dtype)\n"
+    }
+  ]
+}
diff --git a/solutions/dense_attention/sglang_noncausal_dense_attention_h32.json b/solutions/dense_attention/sglang_noncausal_dense_attention_h32.json
deleted file mode 100644
index fe9e6d1778d82a67df625bc93c9d9df0390aba20..0000000000000000000000000000000000000000
--- a/solutions/dense_attention/sglang_noncausal_dense_attention_h32.json
+++ /dev/null
@@ -1,23 +0,0 @@
-{
-  "name": "sglang_noncausal_dense_attention_h32",
-  "definition": "noncausal_dense_attention_h32",
-  "author": "kda",
-  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B300"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for noncausal_dense_attention_h32: LTX-2.5 audio (D=64) and\nvideo (D=128) attention.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: LTX2Attention.forward\n(models/dits/ltx_2.py:997-1006) calls ``self.attn(q, k, v, attn_mask=mask)``\non [B, N, H, D] Q/K/V for the video ``attn1`` (USPAttention, :857-871) and\n``attn2`` (LocalAttention, :841-855) and the audio ``audio_attn1``,\n``audio_to_video_attn`` and ``video_to_audio_attn`` (USPAttention or\nLocalAttention) and ``audio_attn2`` (LocalAttention); every layer has\nsoftmax_scale=None -> head_size**-0.5 and allow_cudnn_sdp=True. On one GPU\nthe mask is None, except for the prompt cross attentions' all-ones mask,\nwhose additive mask is exactly zero; the definition has no mask.\n\nUSPAttention.forward (layers/attention/layer.py:1466-1468) or\nLocalAttention.forward (:799) -> SDPAImpl.forward\n(layers/attention/backends/sdpa.py:76-114): one SDPA call under\n``sdpa_kernel(_PYTORCH_DEFAULT_CUDA_SDP_BACKENDS, set_priority=True)``, which\ntries cuDNN, Flash and memory-efficient attention before math.\n\nSDPAImpl returns a [B, Nq, H, D] transpose of the SDPA output;\n``.contiguous().detach()`` hands it over dense, without a copy when the kernel\nwrote it sequence-major.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.attention.backends.sdpa import SDPAImpl\n\n\ndef prepare(Q, K, V, sm_scale):\n    impl = SDPAImpl(\n        num_heads=Q.shape[2],\n        head_size=Q.shape[3],\n        causal=False,\n        softmax_scale=sm_scale,\n        num_kv_heads=K.shape[2],\n        allow_cudnn_sdp=True,\n    )\n\n    def launch():\n        return impl.forward(Q, K, V, None).contiguous().detach()\n\n    return launch\n\n\ndef run(Q, K, V, sm_scale):\n    return prepare(Q, K, V, sm_scale)()\n"
-    }
-  ]
-}
diff --git a/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d128.json b/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d128.json
new file mode 100644
index 0000000000000000000000000000000000000000..c52c16e69ee6de0ed92ba5e9813477f0c1be79df
--- /dev/null
+++ b/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d128.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_noncausal_dense_attention_h32_d128",
+  "definition": "noncausal_dense_attention_h32_d128",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for noncausal_dense_attention_h32_d128: LTX-2.5 video attention.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: LTX2Attention.forward\n(models/dits/ltx_2.py:997-1006) calls ``self.attn(q, k, v, attn_mask=mask)``\non [B, N, H, D] Q/K/V for ``attn1`` (USPAttention, :857-871) and ``attn2``\n(LocalAttention, :841-855); every layer has\nsoftmax_scale=None -> head_size**-0.5 and allow_cudnn_sdp=True. On one GPU the\nmask is None, except for ``attn2``'s all-ones prompt mask, whose additive\nmask is exactly zero; the definition has no mask.\n\nUSPAttention.forward (layers/attention/layer.py:1466-1468) or\nLocalAttention.forward (:799) -> SDPAImpl.forward\n(layers/attention/backends/sdpa.py:76-114): one SDPA call under\n``sdpa_kernel(_PYTORCH_DEFAULT_CUDA_SDP_BACKENDS, set_priority=True)``, which\ntries cuDNN, Flash and memory-efficient attention before math.\n\nSDPAImpl returns a [B, Nq, H, D] transpose of the SDPA output;\n``.contiguous().detach()`` hands it over dense, without a copy when the kernel\nwrote it sequence-major.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.attention.backends.sdpa import SDPAImpl\n\n\ndef prepare(Q, K, V, sm_scale):\n    impl = SDPAImpl(\n        num_heads=Q.shape[2],\n        head_size=Q.shape[3],\n        causal=False,\n        softmax_scale=sm_scale,\n        num_kv_heads=K.shape[2],\n        allow_cudnn_sdp=True,\n    )\n\n    def launch():\n        return impl.forward(Q, K, V, None).contiguous().detach()\n\n    return launch\n\n\ndef run(Q, K, V, sm_scale):\n    return prepare(Q, K, V, sm_scale)()\n"
+    }
+  ]
+}
diff --git a/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d64.json b/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d64.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8da2b944acabada715ac87e0b14d36327f890ed
--- /dev/null
+++ b/solutions/dense_attention/sglang_noncausal_dense_attention_h32_d64.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_noncausal_dense_attention_h32_d64",
+  "definition": "noncausal_dense_attention_h32_d64",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for noncausal_dense_attention_h32_d64: LTX-2.5 audio attention.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: LTX2Attention.forward\n(models/dits/ltx_2.py:997-1006) calls ``self.attn(q, k, v, attn_mask=mask)``\non [B, N, H, D] Q/K/V for ``audio_attn1``, ``audio_to_video_attn`` and\n``video_to_audio_attn`` (USPAttention, :857-871, or LocalAttention) and\n``audio_attn2`` (LocalAttention, :841-855); every layer has\nsoftmax_scale=None -> head_size**-0.5 and allow_cudnn_sdp=True. On one GPU the\nmask is None, except for ``audio_attn2``'s all-ones prompt mask, whose additive\nmask is exactly zero; the definition has no mask.\n\nUSPAttention.forward (layers/attention/layer.py:1466-1468) or\nLocalAttention.forward (:799) -> SDPAImpl.forward\n(layers/attention/backends/sdpa.py:76-114): one SDPA call under\n``sdpa_kernel(_PYTORCH_DEFAULT_CUDA_SDP_BACKENDS, set_priority=True)``, which\ntries cuDNN, Flash and memory-efficient attention before math.\n\nSDPAImpl returns a [B, Nq, H, D] transpose of the SDPA output;\n``.contiguous().detach()`` hands it over dense, without a copy when the kernel\nwrote it sequence-major.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.attention.backends.sdpa import SDPAImpl\n\n\ndef prepare(Q, K, V, sm_scale):\n    impl = SDPAImpl(\n        num_heads=Q.shape[2],\n        head_size=Q.shape[3],\n        causal=False,\n        softmax_scale=sm_scale,\n        num_kv_heads=K.shape[2],\n        allow_cudnn_sdp=True,\n    )\n\n    def launch():\n        return impl.forward(Q, K, V, None).contiguous().detach()\n\n    return launch\n\n\ndef run(Q, K, V, sm_scale):\n    return prepare(Q, K, V, sm_scale)()\n"
+    }
+  ]
+}
diff --git a/solutions/gemm/deepgemm_bf16.json b/solutions/gemm/deepgemm_bf16.json
deleted file mode 100644
index 6b29ac0f65f42f8c9664e16b4741e8df6cc59c07..0000000000000000000000000000000000000000
--- a/solutions/gemm/deepgemm_bf16.json
+++ /dev/null
@@ -1,24 +0,0 @@
-{
-  "name": "deepgemm_bf16",
-  "definition": "dense_gemm_bf16_nt",
-  "author": "inference-kernels",
-  "description": "DeepGEMM Blackwell BF16 GEMM with FP32 accumulation, for every (N, K). Selected by local B200 comparison against FlashInfer CUTLASS, cuTile, and small-row CuTe DSL; the CuTe DSL branch for 1-4 or 32 rows went with those rows. Merged by KDA from the per-shape solutions when the per-shape definitions were merged. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B200"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "torch>=2.11",
-      "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "import torch\nimport deep_gemm\n\ndef run(A, W):\n    output = torch.empty((A.shape[0], W.shape[0]), dtype=torch.bfloat16, device=A.device)\n    deep_gemm.bf16_gemm_nt(A, W, output)\n    return output\n"
-    }
-  ]
-}
diff --git a/solutions/gemm/deepgemm_bf16_n12288_k4096.json b/solutions/gemm/deepgemm_bf16_n12288_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1200e57b64adbe807576f292a122960b0738a77
--- /dev/null
+++ b/solutions/gemm/deepgemm_bf16_n12288_k4096.json
@@ -0,0 +1,8 @@
+{
+  "name": "deepgemm_bf16_n12288_k4096",
+  "definition": "dense_gemm_bf16_nt_n12288_k4096",
+  "author": "inference-kernels",
+  "description": "DeepGEMM Blackwell BF16 GEMM with FP32 accumulation. Selected by local B200 comparison against FlashInfer CUTLASS, cuTile, and small-row CuTe DSL. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "import torch\nimport deep_gemm\n\ndef run(A, W):\n    output = torch.empty((A.shape[0], W.shape[0]), dtype=torch.bfloat16, device=A.device)\n    deep_gemm.bf16_gemm_nt(A, W, output)\n    return output\n"}]
+}
diff --git a/solutions/gemm/deepgemm_bf16_n4096_k12288.json b/solutions/gemm/deepgemm_bf16_n4096_k12288.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b5660eda40a624135e4eb7383adb8f4a59b8118
--- /dev/null
+++ b/solutions/gemm/deepgemm_bf16_n4096_k12288.json
@@ -0,0 +1,8 @@
+{
+  "name": "deepgemm_bf16_n4096_k12288",
+  "definition": "dense_gemm_bf16_nt_n4096_k12288",
+  "author": "inference-kernels",
+  "description": "DeepGEMM Blackwell BF16 GEMM with FP32 accumulation. Selected by local B200 comparison against FlashInfer CUTLASS, cuTile, and small-row CuTe DSL. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "import torch\nimport deep_gemm\n\ndef run(A, W):\n    output = torch.empty((A.shape[0], W.shape[0]), dtype=torch.bfloat16, device=A.device)\n    deep_gemm.bf16_gemm_nt(A, W, output)\n    return output\n"}]
+}
diff --git a/solutions/gemm/hybrid_bf16_n4096_k4096.json b/solutions/gemm/hybrid_bf16_n4096_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e03ae0ebb992f6ab2bcf9f1f8b0d36b389b237c
--- /dev/null
+++ b/solutions/gemm/hybrid_bf16_n4096_k4096.json
@@ -0,0 +1,8 @@
+{
+  "name": "hybrid_bf16_n4096_k4096",
+  "definition": "dense_gemm_bf16_nt_n4096_k4096",
+  "author": "inference-kernels",
+  "description": "Shape-based B200 dispatch between FlashInfer CuTe DSL for 1-4 or 32 rows and DeepGEMM otherwise. Selected by local comparisons on actual captured shapes. Upstreams: https://github.com/flashinfer-ai/flashinfer (Apache-2.0), https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "flashinfer-python==0.7.2", "nvidia-cutlass-dsl[cu13]==4.7.1", "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "import torch\nimport deep_gemm\nfrom flashinfer.gemm import mm_bf16\n\ndef run(A, W):\n    rows = A.shape[0]\n    if rows <= 4 or rows == 32:\n        return mm_bf16(A, W.T, backend='cute-dsl')\n    output = torch.empty((rows, W.shape[0]), dtype=torch.bfloat16, device=A.device)\n    deep_gemm.bf16_gemm_nt(A, W, output)\n    return output\n"}]
+}
diff --git a/solutions/gemm/sglang_dense_gemm_bf16_nt.json b/solutions/gemm/sglang_dense_gemm_bf16_nt.json
deleted file mode 100644
index ac87a627b8a0a05812804f554655fa865a593cdf..0000000000000000000000000000000000000000
--- a/solutions/gemm/sglang_dense_gemm_bf16_nt.json
+++ /dev/null
@@ -1,23 +0,0 @@
-{
-  "name": "sglang_dense_gemm_bf16_nt",
-  "definition": "dense_gemm_bf16_nt",
-  "author": "kda",
-  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B300"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for dense_gemm_bf16_nt: C = A @ W.T in BF16.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/. Every Qwen-Image-2.1 projection below\nends in UnquantizedLinearMethod.apply (layers/linear.py:203) ->\napply_unquantized_linear(x, weight, None) (:155) -> F.linear(x, weight, None)\n(:169):\n- (N, K) = (4096, 4096): ``attn.to_q`` ColumnParallelLinear\n  (models/dits/qwen_image21.py:369, called at :487/:723) and ``attn.to_out.0``\n  RowParallelLinear (:380, called at :741), via layers/linear.py:518/:1185.\n  The nn.Linear layers ``txt_in.in_layer`` (:294) and\n  ``timestep_embedder.linear_2`` (:308) issue the same F.linear call.\n- (12288, 4096): ``img_mlp.gate_layer`` ColumnParallelLinear (:331, called at\n  :347), via layers/linear.py:518. The batch-1 packed Q/K/V projection,\n  torch.mm(x.view(seq, C), packed.t(), out=buffer[0, prefix:]) (:579), gives\n  the same bytes.\n- (4096, 12288): ``img_mlp.out`` RowParallelLinear (:338, called at\n  :356/:360), via layers/linear.py:1185.\n\nAdapter: none. A is SGLang's x as [M, K] rows (the engine's [1, M, K]\nactivation gives the same bytes), W is layer.weight, bias is None. The\nworker's numerics policy (utils/numerics_policy.py:33, applied at\nmanagers/gpu_worker.py:420 with the server_args.py:510-511 defaults) keeps\nPyTorch's defaults, so nothing is set here.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.linear import apply_unquantized_linear\n\n\ndef run(A, W):\n    return apply_unquantized_linear(A, W, None)\n"
-    }
-  ]
-}
diff --git a/solutions/gemm/sglang_dense_gemm_bf16_nt_n12288_k4096.json b/solutions/gemm/sglang_dense_gemm_bf16_nt_n12288_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..134fc9835c310d98025af4e8114786fec1bd1cdf
--- /dev/null
+++ b/solutions/gemm/sglang_dense_gemm_bf16_nt_n12288_k4096.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_dense_gemm_bf16_nt_n12288_k4096",
+  "definition": "dense_gemm_bf16_nt_n12288_k4096",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for dense_gemm_bf16_nt_n12288_k4096: C = A @ W.T in BF16.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: the Qwen-Image-2.1 ``img_mlp.gate_layer``\nColumnParallelLinear (models/dits/qwen_image21.py:331, called at :347) ->\nColumnParallelLinear.forward (layers/linear.py:518) ->\nUnquantizedLinearMethod.apply (:203) -> apply_unquantized_linear(x, weight,\nNone) (:155) -> F.linear(x, weight, None) (:169). The batch-1 packed Q/K/V\nprojection, torch.mm(x.view(seq, C), packed.t(), out=buffer[0, prefix:])\n(models/dits/qwen_image21.py:579), gives the same bytes.\n\nAdapter: none. A is SGLang's x as [M, K] rows (the engine's [1, M, K]\nactivation gives the same bytes), W is layer.weight, bias is None. The\nworker's numerics policy (utils/numerics_policy.py:33, applied at\nmanagers/gpu_worker.py:420 with the server_args.py:510-511 defaults) keeps\nPyTorch's defaults, so nothing is set here.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.linear import apply_unquantized_linear\n\n\ndef run(A, W):\n    return apply_unquantized_linear(A, W, None)\n"
+    }
+  ]
+}
diff --git a/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k12288.json b/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k12288.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa60eef32f6c5cbb260fa670bb699202d9f71be6
--- /dev/null
+++ b/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k12288.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_dense_gemm_bf16_nt_n4096_k12288",
+  "definition": "dense_gemm_bf16_nt_n4096_k12288",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for dense_gemm_bf16_nt_n4096_k12288: C = A @ W.T in BF16.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: the Qwen-Image-2.1 ``img_mlp.out``\nRowParallelLinear (models/dits/qwen_image21.py:338, called at :356/:360) ->\nRowParallelLinear.forward (layers/linear.py:1185) ->\nUnquantizedLinearMethod.apply (:203) -> apply_unquantized_linear(x, weight,\nNone) (:155) -> F.linear(x, weight, None) (:169).\n\nAdapter: none. A is SGLang's x as [M, K] rows (the engine's [1, M, K]\nactivation gives the same bytes), W is layer.weight, bias is None. The\nworker's numerics policy (utils/numerics_policy.py:33, applied at\nmanagers/gpu_worker.py:420 with the server_args.py:510-511 defaults) keeps\nPyTorch's defaults, so nothing is set here.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.linear import apply_unquantized_linear\n\n\ndef run(A, W):\n    return apply_unquantized_linear(A, W, None)\n"
+    }
+  ]
+}
diff --git a/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k4096.json b/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k4096.json
new file mode 100644
index 0000000000000000000000000000000000000000..61da369d12664e7f6822781e3ca3beeb57da8656
--- /dev/null
+++ b/solutions/gemm/sglang_dense_gemm_bf16_nt_n4096_k4096.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_dense_gemm_bf16_nt_n4096_k4096",
+  "definition": "dense_gemm_bf16_nt_n4096_k4096",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for dense_gemm_bf16_nt_n4096_k4096: C = A @ W.T in BF16.\n\nSGLang diffusion runtime at commit 684fea8a, paths under\npython/sglang/multimodal_gen/runtime/: the Qwen-Image-2.1 ``attn.to_q``\nColumnParallelLinear (models/dits/qwen_image21.py:369, called at :487/:723)\nand ``attn.to_out.0`` RowParallelLinear (:380, called at :741) ->\nlayers/linear.py:518/:1185 -> UnquantizedLinearMethod.apply (:203) ->\napply_unquantized_linear(x, weight, None) (:155) -> F.linear(x, weight, None)\n(:169). The nn.Linear layers ``txt_in.in_layer`` (models/dits/qwen_image21.py:294)\nand ``timestep_embedder.linear_2`` (:308) issue the same F.linear call.\n\nAdapter: none. A is SGLang's x as [M, K] rows (the engine's batched\nactivation gives the same bytes), W is layer.weight, bias is None. The\nworker's numerics policy (utils/numerics_policy.py:33, applied at\nmanagers/gpu_worker.py:420 with the server_args.py:510-511 defaults) keeps\nPyTorch's defaults, so nothing is set here.\n\"\"\"\n\nfrom sglang.multimodal_gen.runtime.layers.linear import apply_unquantized_linear\n\n\ndef run(A, W):\n    return apply_unquantized_linear(A, W, None)\n"
+    }
+  ]
+}
diff --git a/solutions/grouped_gemm/deepgemm_fp8_fp4.json b/solutions/grouped_gemm/deepgemm_fp8_fp4.json
deleted file mode 100644
index d47dc044f0aa419f4ce13c8d4ea619c5dd22e72c..0000000000000000000000000000000000000000
--- a/solutions/grouped_gemm/deepgemm_fp8_fp4.json
+++ /dev/null
@@ -1,24 +0,0 @@
-{
-  "name": "deepgemm_fp8_fp4",
-  "definition": "grouped_gemm_fp8_fp4_e384_g32",
-  "author": "inference-kernels",
-  "description": "DeepGEMM MXFP8 x MXFP4 grouped GEMM. Aligns expert segments, packs scales without requantization, and restores original row order. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT). Merged by KDA from the per-shape solutions when the per-shape definitions were merged.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B200"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "torch>=2.11",
-      "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "import torch\nimport deep_gemm\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    experts, n, _ = Wq_packed.shape\n    alignment = deep_gemm.get_mk_alignment_for_contiguous_layout()\n    capacity = ((m + min(m, experts) * (alignment - 1) + alignment - 1) // alignment) * alignment\n    boundaries = torch.searchsorted(expert_id, torch.arange(experts + 1, device=Aq.device, dtype=torch.int32))\n    counts = boundaries[1:] - boundaries[:-1]\n    padded_counts = ((counts + alignment - 1) // alignment) * alignment\n    offsets = padded_counts.cumsum(0) - padded_counts\n    ids = expert_id.long()\n    row_map = offsets[ids] + torch.arange(m, device=Aq.device) - boundaries[ids]\n    a = torch.zeros((capacity, k), device=Aq.device, dtype=Aq.dtype)\n    a.view(torch.uint8).index_copy_(0, row_map, Aq.view(torch.uint8))\n    sa = torch.full((capacity, k // 32), 127, device=Aq.device, dtype=torch.int8)\n    sa.index_copy_(0, row_map, scale_A)\n    sa = sa.view(torch.int32).t().contiguous().t()\n    sw = scale_W.view(torch.int32).transpose(-1, -2).contiguous().transpose(-1, -2)\n    layout = torch.full((capacity,), -1, device=Aq.device, dtype=torch.int32)\n    layout.index_copy_(0, row_map, expert_id)\n    output = torch.empty((capacity, n), device=Aq.device, dtype=torch.bfloat16)\n    deep_gemm.m_grouped_fp8_fp4_gemm_nt_contiguous(\n        (a, sa), (Wq_packed, sw), output, layout,\n        recipe_a=(1, 32), recipe_b=(1, 32),\n    )\n    return output.index_select(0, row_map)\n"
-    }
-  ]
-}
diff --git a/solutions/grouped_gemm/deepgemm_n4608_k5120.json b/solutions/grouped_gemm/deepgemm_n4608_k5120.json
new file mode 100644
index 0000000000000000000000000000000000000000..28e0db26c95a343aae319334f13355fa939d3658
--- /dev/null
+++ b/solutions/grouped_gemm/deepgemm_n4608_k5120.json
@@ -0,0 +1,8 @@
+{
+  "name": "deepgemm_n4608_k5120",
+  "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32",
+  "author": "inference-kernels",
+  "description": "DeepGEMM MXFP8 x MXFP4 grouped GEMM. Aligns expert segments, packs scales without requantization, and restores original row order. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "import torch\nimport deep_gemm\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    experts, n, _ = Wq_packed.shape\n    alignment = deep_gemm.get_mk_alignment_for_contiguous_layout()\n    capacity = ((m + min(m, experts) * (alignment - 1) + alignment - 1) // alignment) * alignment\n    boundaries = torch.searchsorted(expert_id, torch.arange(experts + 1, device=Aq.device, dtype=torch.int32))\n    counts = boundaries[1:] - boundaries[:-1]\n    padded_counts = ((counts + alignment - 1) // alignment) * alignment\n    offsets = padded_counts.cumsum(0) - padded_counts\n    ids = expert_id.long()\n    row_map = offsets[ids] + torch.arange(m, device=Aq.device) - boundaries[ids]\n    a = torch.zeros((capacity, k), device=Aq.device, dtype=Aq.dtype)\n    a.view(torch.uint8).index_copy_(0, row_map, Aq.view(torch.uint8))\n    sa = torch.full((capacity, k // 32), 127, device=Aq.device, dtype=torch.int8)\n    sa.index_copy_(0, row_map, scale_A)\n    sa = sa.view(torch.int32).t().contiguous().t()\n    sw = scale_W.view(torch.int32).transpose(-1, -2).contiguous().transpose(-1, -2)\n    layout = torch.full((capacity,), -1, device=Aq.device, dtype=torch.int32)\n    layout.index_copy_(0, row_map, expert_id)\n    output = torch.empty((capacity, n), device=Aq.device, dtype=torch.bfloat16)\n    deep_gemm.m_grouped_fp8_fp4_gemm_nt_contiguous(\n        (a, sa), (Wq_packed, sw), output, layout,\n        recipe_a=(1, 32), recipe_b=(1, 32),\n    )\n    return output.index_select(0, row_map)\n"}]
+}
diff --git a/solutions/grouped_gemm/deepgemm_n5120_k2304.json b/solutions/grouped_gemm/deepgemm_n5120_k2304.json
new file mode 100644
index 0000000000000000000000000000000000000000..755f86c0ecdeb0ee4a07fd869ba398ebf03a49ba
--- /dev/null
+++ b/solutions/grouped_gemm/deepgemm_n5120_k2304.json
@@ -0,0 +1,8 @@
+{
+  "name": "deepgemm_n5120_k2304",
+  "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32",
+  "author": "inference-kernels",
+  "description": "DeepGEMM MXFP8 x MXFP4 grouped GEMM. Aligns expert segments, packs scales without requantization, and restores original row order. Upstream: https://github.com/deepseek-ai/DeepGEMM/tree/1fc9c66975ee322d628b3288534ca0e9ce12a64d (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "deep-gemm @ git+https://github.com/deepseek-ai/DeepGEMM.git@1fc9c66975ee322d628b3288534ca0e9ce12a64d"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "import torch\nimport deep_gemm\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    experts, n, _ = Wq_packed.shape\n    alignment = deep_gemm.get_mk_alignment_for_contiguous_layout()\n    capacity = ((m + min(m, experts) * (alignment - 1) + alignment - 1) // alignment) * alignment\n    boundaries = torch.searchsorted(expert_id, torch.arange(experts + 1, device=Aq.device, dtype=torch.int32))\n    counts = boundaries[1:] - boundaries[:-1]\n    padded_counts = ((counts + alignment - 1) // alignment) * alignment\n    offsets = padded_counts.cumsum(0) - padded_counts\n    ids = expert_id.long()\n    row_map = offsets[ids] + torch.arange(m, device=Aq.device) - boundaries[ids]\n    a = torch.zeros((capacity, k), device=Aq.device, dtype=Aq.dtype)\n    a.view(torch.uint8).index_copy_(0, row_map, Aq.view(torch.uint8))\n    sa = torch.full((capacity, k // 32), 127, device=Aq.device, dtype=torch.int8)\n    sa.index_copy_(0, row_map, scale_A)\n    sa = sa.view(torch.int32).t().contiguous().t()\n    sw = scale_W.view(torch.int32).transpose(-1, -2).contiguous().transpose(-1, -2)\n    layout = torch.full((capacity,), -1, device=Aq.device, dtype=torch.int32)\n    layout.index_copy_(0, row_map, expert_id)\n    output = torch.empty((capacity, n), device=Aq.device, dtype=torch.bfloat16)\n    deep_gemm.m_grouped_fp8_fp4_gemm_nt_contiguous(\n        (a, sa), (Wq_packed, sw), output, layout,\n        recipe_a=(1, 32), recipe_b=(1, 32),\n    )\n    return output.index_select(0, row_map)\n"}]
+}
diff --git a/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_g32.json b/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_g32.json
deleted file mode 100644
index 7645e0038f05b0ec0a932375dd05dfd0d751fb4d..0000000000000000000000000000000000000000
--- a/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_g32.json
+++ /dev/null
@@ -1,23 +0,0 @@
-{
-  "name": "sglang_grouped_gemm_fp8_fp4_e384_g32",
-  "definition": "grouped_gemm_fp8_fp4_e384_g32",
-  "author": "kda",
-  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B300"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang's DeepGEMM MoE expert GEMMs: the fused gate/up projection (N=4608,\nK=5120) and the down projection (N=5120, K=2304).\n\nSGLang 684fea8a (paths under ``python/sglang/``), captured with ``--tp-size 4\n--ep-size 4 --moe-runner-backend deep_gemm --moe-a2a-backend none``: each of\nthe 4 EP ranks holds 96 of the 384 experts and runs, for its rows,\n\n- once at load: ``Fp8MoEMethod.process_weights_after_loading_block_quant``\n  packs the fp32 group-32 weight scales with\n  ``deep_gemm.transform_sf_into_required_layout(recipe=(1, 32),\n  num_groups=96)`` (srt/layers/quantization/fp8.py:2132-2147);\n- per batch: ``pre_permute_standard_to_deep_gemm``\n  (srt/layers/moe/moe_runner/deep_gemm.py:988) sizes the compact contiguous\n  layout (``get_contiguous_layout_alignment`` :1082, ``_get_compact_all_tokens``\n  :1085, ``pad_expert_counts`` :1093) and ``ep_scatter`` (:1166) writes\n  ``m_indices`` (-1 on padding rows) and, for gate/up, the FP8 rows and the\n  MN-major packed UE8M0 scales (:1149); for the down projection\n  ``_run_contiguous_gemm`` quantizes the gate/up activation into the same\n  padded rows. ``DeepGemmRunnerCore.run`` enters\n  ``contiguous_layout_alignment_scope(block_e)`` (:378) and\n  ``_run_contiguous_gemm`` calls (gate/up :456, down :631)\n  ``deep_gemm_wrapper.grouped_gemm_nt_f8f8bf16_contig``\n  (srt/layers/deep_gemm_wrapper/entrypoint.py:132), i.e.\n  ``deep_gemm.m_grouped_fp8_gemm_nt_contiguous`` (:157), with recipe_a =\n  recipe_b = (1, 32): the capture's engine_adapter passes the actual\n  activation group instead of the FP4-expert default (1, 128) (:337).\n\nInterface adapter: the definition's rows are already sorted by expert, so\n``prepare`` derives the layout from ``expert_id`` (each rank's padded\nsegments, row destinations, ``m_indices``), allocates the per-rank buffers,\nand does the load-time weight work; ``launch`` scatters the current FP8 rows\nand scale bytes (four UE8M0 bytes per int32) into the padded layout, runs the\nfour rank GEMMs and gathers the valid rows.\n\"\"\"\n\nimport deep_gemm\nimport torch\nfrom sglang.kernels.ops.moe.triton_pad_expert_counts import pad_expert_counts\nfrom sglang.srt.layers.deep_gemm_wrapper.entrypoint import (\n    contiguous_layout_alignment_scope,\n    get_contiguous_layout_alignment,\n)\nfrom sglang.srt.layers.moe.moe_runner.deep_gemm import _get_compact_all_tokens\n\nEP_SIZE = 4  # capture --ep-size\nRECIPE = (1, 32)  # capture engine_adapter: actual activation group\n\n\n@torch.no_grad()\ndef prepare(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    num_experts, n, _ = Wq_packed.shape\n    local = num_experts // EP_SIZE\n    weights = [\n        (\n            Wq_packed[e0 : e0 + local],\n            deep_gemm.transform_sf_into_required_layout(\n                torch.exp2(scale_W[e0 : e0 + local].view(torch.uint8).float() - 127),\n                mn=n,\n                k=k,\n                recipe=RECIPE,\n                num_groups=local,\n                disable_ue8m0_cast=False,\n            ),\n        )\n        for e0 in range(0, num_experts, local)\n    ]\n\n    block_e = get_contiguous_layout_alignment(m, local)\n    all_tokens = _get_compact_all_tokens(m, local, block_e)\n    eid = expert_id.long()\n    counts = torch.bincount(eid, minlength=num_experts)\n    padded = torch.stack(\n        [pad_expert_counts(c, block_e, all_tokens) for c in counts.view(-1, local)]\n    ).flatten()\n    rank, local_eid = eid // local, eid % local\n    dst = (\n        (torch.cumsum(padded, 0) - padded)[eid]\n        + torch.arange(m, device=eid.device)\n        - (torch.cumsum(counts, 0) - counts)[eid]\n    )\n    m_indices = torch.full(\n        (EP_SIZE * all_tokens,), -1, dtype=torch.int32, device=Aq.device\n    )\n    m_indices[dst] = local_eid.int()\n    a = torch.zeros(EP_SIZE * all_tokens, k, dtype=Aq.dtype, device=Aq.device)\n    sfa = torch.zeros(\n        EP_SIZE, k // 128, all_tokens, dtype=torch.int32, device=Aq.device\n    )\n    out = torch.empty(EP_SIZE * all_tokens, n, dtype=torch.bfloat16, device=Aq.device)\n    y = torch.empty(m, n, dtype=torch.bfloat16, device=Aq.device)\n    rows = [slice(r * all_tokens, (r + 1) * all_tokens) for r in range(EP_SIZE)]\n    local_dst = dst - rank * all_tokens\n\n    def launch():\n        a.view(torch.uint8).index_copy_(0, dst, Aq.view(torch.uint8))\n        sfa.transpose(1, 2)[rank, local_dst] = scale_A.view(torch.int32)\n        with contiguous_layout_alignment_scope(block_e):\n            for r, (w, sfw) in enumerate(weights):\n                deep_gemm.m_grouped_fp8_gemm_nt_contiguous(\n                    (a[rows[r]], sfa[r].T),\n                    (w, sfw),\n                    out[rows[r]],\n                    m_indices[rows[r]],\n                    recipe_a=RECIPE,\n                    recipe_b=RECIPE,\n                )\n        torch.index_select(out, 0, dst, out=y)\n        return y\n\n    return launch\n\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    return prepare(Aq, Wq_packed, scale_A, scale_W, expert_id)()\n"
-    }
-  ]
-}
diff --git a/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json b/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
new file mode 100644
index 0000000000000000000000000000000000000000..26d19dcba5ea19d281192af1fbdcb457585983eb
--- /dev/null
+++ b/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_grouped_gemm_fp8_fp4_e384_n4608_k5120_g32",
+  "definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang's DeepGEMM MoE expert GEMM, fused gate/up projection (N=4608, K=5120).\n\nSGLang 684fea8a (paths under ``python/sglang/``), captured with ``--tp-size 4\n--ep-size 4 --moe-runner-backend deep_gemm --moe-a2a-backend none``: each of\nthe 4 EP ranks holds 96 of the 384 experts and runs, for its rows,\n\n- once at load: ``Fp8MoEMethod.process_weights_after_loading_block_quant``\n  packs the fp32 group-32 weight scales with\n  ``deep_gemm.transform_sf_into_required_layout(recipe=(1, 32),\n  num_groups=96)`` (srt/layers/quantization/fp8.py:2132-2147);\n- per batch: ``pre_permute_standard_to_deep_gemm``\n  (srt/layers/moe/moe_runner/deep_gemm.py:988) sizes the compact contiguous\n  layout (``get_contiguous_layout_alignment`` :1082, ``_get_compact_all_tokens``\n  :1085, ``pad_expert_counts`` :1093) and ``ep_scatter`` (:1166) writes the FP8\n  rows, the MN-major packed UE8M0 scales (:1149) and ``m_indices`` (-1 on\n  padding rows); ``DeepGemmRunnerCore.run`` enters\n  ``contiguous_layout_alignment_scope(block_e)`` (:378) and\n  ``_run_contiguous_gemm`` (:456) calls\n  ``deep_gemm_wrapper.grouped_gemm_nt_f8f8bf16_contig``\n  (srt/layers/deep_gemm_wrapper/entrypoint.py:132), i.e.\n  ``deep_gemm.m_grouped_fp8_gemm_nt_contiguous`` (:157), with recipe_a =\n  recipe_b = (1, 32): the capture's engine_adapter passes the actual\n  activation group instead of the FP4-expert default (1, 128) (:337).\n\nInterface adapter: the definition's rows are already sorted by expert, so\n``prepare`` derives the layout from ``expert_id`` (each rank's padded\nsegments, row destinations, ``m_indices``), allocates the per-rank buffers,\nand does the load-time weight work; ``launch`` scatters the current FP8 rows\nand scale bytes (four UE8M0 bytes per int32) into the padded layout, runs the\nfour rank GEMMs and gathers the valid rows.\n\"\"\"\n\nimport deep_gemm\nimport torch\nfrom sglang.kernels.ops.moe.triton_pad_expert_counts import pad_expert_counts\nfrom sglang.srt.layers.deep_gemm_wrapper.entrypoint import (\n    contiguous_layout_alignment_scope,\n    get_contiguous_layout_alignment,\n)\nfrom sglang.srt.layers.moe.moe_runner.deep_gemm import _get_compact_all_tokens\n\nEP_SIZE = 4  # capture --ep-size\nRECIPE = (1, 32)  # capture engine_adapter: actual activation group\n\n\n@torch.no_grad()\ndef prepare(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    num_experts, n, _ = Wq_packed.shape\n    local = num_experts // EP_SIZE\n    weights = [\n        (\n            Wq_packed[e0 : e0 + local],\n            deep_gemm.transform_sf_into_required_layout(\n                torch.exp2(scale_W[e0 : e0 + local].view(torch.uint8).float() - 127),\n                mn=n,\n                k=k,\n                recipe=RECIPE,\n                num_groups=local,\n                disable_ue8m0_cast=False,\n            ),\n        )\n        for e0 in range(0, num_experts, local)\n    ]\n\n    block_e = get_contiguous_layout_alignment(m, local)\n    all_tokens = _get_compact_all_tokens(m, local, block_e)\n    eid = expert_id.long()\n    counts = torch.bincount(eid, minlength=num_experts)\n    padded = torch.stack(\n        [pad_expert_counts(c, block_e, all_tokens) for c in counts.view(-1, local)]\n    ).flatten()\n    rank, local_eid = eid // local, eid % local\n    dst = (\n        (torch.cumsum(padded, 0) - padded)[eid]\n        + torch.arange(m, device=eid.device)\n        - (torch.cumsum(counts, 0) - counts)[eid]\n    )\n    m_indices = torch.full(\n        (EP_SIZE * all_tokens,), -1, dtype=torch.int32, device=Aq.device\n    )\n    m_indices[dst] = local_eid.int()\n    a = torch.zeros(EP_SIZE * all_tokens, k, dtype=Aq.dtype, device=Aq.device)\n    sfa = torch.zeros(\n        EP_SIZE, k // 128, all_tokens, dtype=torch.int32, device=Aq.device\n    )\n    out = torch.empty(EP_SIZE * all_tokens, n, dtype=torch.bfloat16, device=Aq.device)\n    y = torch.empty(m, n, dtype=torch.bfloat16, device=Aq.device)\n    rows = [slice(r * all_tokens, (r + 1) * all_tokens) for r in range(EP_SIZE)]\n    local_dst = dst - rank * all_tokens\n\n    def launch():\n        a.view(torch.uint8).index_copy_(0, dst, Aq.view(torch.uint8))\n        sfa.transpose(1, 2)[rank, local_dst] = scale_A.view(torch.int32)\n        with contiguous_layout_alignment_scope(block_e):\n            for r, (w, sfw) in enumerate(weights):\n                deep_gemm.m_grouped_fp8_gemm_nt_contiguous(\n                    (a[rows[r]], sfa[r].T),\n                    (w, sfw),\n                    out[rows[r]],\n                    m_indices[rows[r]],\n                    recipe_a=RECIPE,\n                    recipe_b=RECIPE,\n                )\n        torch.index_select(out, 0, dst, out=y)\n        return y\n\n    return launch\n\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    return prepare(Aq, Wq_packed, scale_A, scale_W, expert_id)()\n"
+    }
+  ]
+}
diff --git a/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json b/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f6fb100247e564c27ca0d4da3d1eac67df5eb1f
--- /dev/null
+++ b/solutions/grouped_gemm/sglang_grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_grouped_gemm_fp8_fp4_e384_n5120_k2304_g32",
+  "definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang's DeepGEMM MoE expert GEMM, down projection (N=5120, K=2304).\n\nSGLang 684fea8a (paths under ``python/sglang/``), captured with ``--tp-size 4\n--ep-size 4 --moe-runner-backend deep_gemm --moe-a2a-backend none``: each of\nthe 4 EP ranks holds 96 of the 384 experts and runs, for its rows,\n\n- once at load: ``Fp8MoEMethod.process_weights_after_loading_block_quant``\n  packs the fp32 group-32 weight scales with\n  ``deep_gemm.transform_sf_into_required_layout(recipe=(1, 32),\n  num_groups=96)`` (srt/layers/quantization/fp8.py:2132-2147);\n- per batch: ``pre_permute_standard_to_deep_gemm``\n  (srt/layers/moe/moe_runner/deep_gemm.py:988) sizes the compact contiguous\n  layout (``get_contiguous_layout_alignment`` :1082, ``_get_compact_all_tokens``\n  :1085, ``pad_expert_counts`` :1093) and ``ep_scatter`` (:1166) writes\n  ``m_indices`` (-1 on padding rows); ``DeepGemmRunnerCore.run`` enters\n  ``contiguous_layout_alignment_scope(block_e)`` (:378) and\n  ``_run_contiguous_gemm`` quantizes the gate/up activation into the same\n  padded rows (FP8 rows, MN-major packed UE8M0 scales) and, for the down\n  projection, calls (:631) ``deep_gemm_wrapper.grouped_gemm_nt_f8f8bf16_contig``\n  (srt/layers/deep_gemm_wrapper/entrypoint.py:132), i.e.\n  ``deep_gemm.m_grouped_fp8_gemm_nt_contiguous`` (:157), with recipe_a =\n  recipe_b = (1, 32): the capture's engine_adapter passes the actual\n  activation group instead of the FP4-expert default (1, 128) (:337).\n\nInterface adapter: the definition's rows are already sorted by expert, so\n``prepare`` derives the layout from ``expert_id`` (each rank's padded\nsegments, row destinations, ``m_indices``), allocates the per-rank buffers,\nand does the load-time weight work; ``launch`` scatters the current FP8 rows\nand scale bytes (four UE8M0 bytes per int32) into the padded layout, runs the\nfour rank GEMMs and gathers the valid rows.\n\"\"\"\n\nimport deep_gemm\nimport torch\nfrom sglang.kernels.ops.moe.triton_pad_expert_counts import pad_expert_counts\nfrom sglang.srt.layers.deep_gemm_wrapper.entrypoint import (\n    contiguous_layout_alignment_scope,\n    get_contiguous_layout_alignment,\n)\nfrom sglang.srt.layers.moe.moe_runner.deep_gemm import _get_compact_all_tokens\n\nEP_SIZE = 4  # capture --ep-size\nRECIPE = (1, 32)  # capture engine_adapter: actual activation group\n\n\n@torch.no_grad()\ndef prepare(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    m, k = Aq.shape\n    num_experts, n, _ = Wq_packed.shape\n    local = num_experts // EP_SIZE\n    weights = [\n        (\n            Wq_packed[e0 : e0 + local],\n            deep_gemm.transform_sf_into_required_layout(\n                torch.exp2(scale_W[e0 : e0 + local].view(torch.uint8).float() - 127),\n                mn=n,\n                k=k,\n                recipe=RECIPE,\n                num_groups=local,\n                disable_ue8m0_cast=False,\n            ),\n        )\n        for e0 in range(0, num_experts, local)\n    ]\n\n    block_e = get_contiguous_layout_alignment(m, local)\n    all_tokens = _get_compact_all_tokens(m, local, block_e)\n    eid = expert_id.long()\n    counts = torch.bincount(eid, minlength=num_experts)\n    padded = torch.stack(\n        [pad_expert_counts(c, block_e, all_tokens) for c in counts.view(-1, local)]\n    ).flatten()\n    rank, local_eid = eid // local, eid % local\n    dst = (\n        (torch.cumsum(padded, 0) - padded)[eid]\n        + torch.arange(m, device=eid.device)\n        - (torch.cumsum(counts, 0) - counts)[eid]\n    )\n    m_indices = torch.full(\n        (EP_SIZE * all_tokens,), -1, dtype=torch.int32, device=Aq.device\n    )\n    m_indices[dst] = local_eid.int()\n    a = torch.zeros(EP_SIZE * all_tokens, k, dtype=Aq.dtype, device=Aq.device)\n    sfa = torch.zeros(\n        EP_SIZE, k // 128, all_tokens, dtype=torch.int32, device=Aq.device\n    )\n    out = torch.empty(EP_SIZE * all_tokens, n, dtype=torch.bfloat16, device=Aq.device)\n    y = torch.empty(m, n, dtype=torch.bfloat16, device=Aq.device)\n    rows = [slice(r * all_tokens, (r + 1) * all_tokens) for r in range(EP_SIZE)]\n    local_dst = dst - rank * all_tokens\n\n    def launch():\n        a.view(torch.uint8).index_copy_(0, dst, Aq.view(torch.uint8))\n        sfa.transpose(1, 2)[rank, local_dst] = scale_A.view(torch.int32)\n        with contiguous_layout_alignment_scope(block_e):\n            for r, (w, sfw) in enumerate(weights):\n                deep_gemm.m_grouped_fp8_gemm_nt_contiguous(\n                    (a[rows[r]], sfa[r].T),\n                    (w, sfw),\n                    out[rows[r]],\n                    m_indices[rows[r]],\n                    recipe_a=RECIPE,\n                    recipe_b=RECIPE,\n                )\n        torch.index_select(out, 0, dst, out=y)\n        return y\n\n    return launch\n\n\ndef run(Aq, Wq_packed, scale_A, scale_W, expert_id):\n    return prepare(Aq, Wq_packed, scale_A, scale_W, expert_id)()\n"
+    }
+  ]
+}
diff --git a/solutions/neighborhood_attention/natten_blackwell.json b/solutions/neighborhood_attention/natten_blackwell.json
deleted file mode 100644
index 41622aac7531950680f94f48f8621f149803c46a..0000000000000000000000000000000000000000
--- a/solutions/neighborhood_attention/natten_blackwell.json
+++ /dev/null
@@ -1,24 +0,0 @@
-{
-  "name": "natten_blackwell",
-  "definition": "neighborhood_attention_3d_d64",
-  "author": "inference-kernels",
-  "description": "NATTEN 0.21.7 Blackwell fused neighborhood attention with inward-shifted boundary windows, for every stage's window. Queries already include the model scale. Merged by KDA from the per-shape solutions when the per-shape definitions were merged. Upstream: https://github.com/SHI-Labs/NATTEN/tree/39375ea1d2f3b1bb284315fa5f79f6ecbebb6382 (MIT).",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B200"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "torch>=2.11",
-      "natten==0.21.7"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "from natten.functional import na3d\n\ndef run(Q, K, V, Kt, Ky, Kx):\n    return na3d(Q, K, V, kernel_size=(int(Kt), int(Ky), int(Kx)), scale=1.0, backend='blackwell-fna')\n"
-    }
-  ]
-}
diff --git a/solutions/neighborhood_attention/natten_blackwell_h16_3x7x7.json b/solutions/neighborhood_attention/natten_blackwell_h16_3x7x7.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd5496b286b0e80354b9faea18b6cc99a3587b91
--- /dev/null
+++ b/solutions/neighborhood_attention/natten_blackwell_h16_3x7x7.json
@@ -0,0 +1,8 @@
+{
+  "name": "natten_blackwell_h16_3x7x7",
+  "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7",
+  "author": "inference-kernels",
+  "description": "NATTEN 0.21.7 Blackwell fused neighborhood attention with inward-shifted boundary windows. Queries already include the model scale. Upstream: https://github.com/SHI-Labs/NATTEN/tree/39375ea1d2f3b1bb284315fa5f79f6ecbebb6382 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "natten==0.21.7"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from natten.functional import na3d\n\ndef run(Q, K, V):\n    return na3d(Q, K, V, kernel_size=(3, 7, 7), scale=1.0, backend='blackwell-fna')\n"}]
+}
diff --git a/solutions/neighborhood_attention/natten_blackwell_h32_3x7x7.json b/solutions/neighborhood_attention/natten_blackwell_h32_3x7x7.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b9622008ca25df88da5c39a15a90de15c0e6790
--- /dev/null
+++ b/solutions/neighborhood_attention/natten_blackwell_h32_3x7x7.json
@@ -0,0 +1,8 @@
+{
+  "name": "natten_blackwell_h32_3x7x7",
+  "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7",
+  "author": "inference-kernels",
+  "description": "NATTEN 0.21.7 Blackwell fused neighborhood attention with inward-shifted boundary windows. Queries already include the model scale. Upstream: https://github.com/SHI-Labs/NATTEN/tree/39375ea1d2f3b1bb284315fa5f79f6ecbebb6382 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "natten==0.21.7"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from natten.functional import na3d\n\ndef run(Q, K, V):\n    return na3d(Q, K, V, kernel_size=(3, 7, 7), scale=1.0, backend='blackwell-fna')\n"}]
+}
diff --git a/solutions/neighborhood_attention/natten_blackwell_h4_11x11x11.json b/solutions/neighborhood_attention/natten_blackwell_h4_11x11x11.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9b1831193ac5f4b2f1dda35da7b4d801b84e3f4
--- /dev/null
+++ b/solutions/neighborhood_attention/natten_blackwell_h4_11x11x11.json
@@ -0,0 +1,8 @@
+{
+  "name": "natten_blackwell_h4_11x11x11",
+  "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11",
+  "author": "inference-kernels",
+  "description": "NATTEN 0.21.7 Blackwell fused neighborhood attention with inward-shifted boundary windows. Queries already include the model scale. Upstream: https://github.com/SHI-Labs/NATTEN/tree/39375ea1d2f3b1bb284315fa5f79f6ecbebb6382 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "natten==0.21.7"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from natten.functional import na3d\n\ndef run(Q, K, V):\n    return na3d(Q, K, V, kernel_size=(11, 11, 11), scale=1.0, backend='blackwell-fna')\n"}]
+}
diff --git a/solutions/neighborhood_attention/natten_blackwell_h8_3x5x5.json b/solutions/neighborhood_attention/natten_blackwell_h8_3x5x5.json
new file mode 100644
index 0000000000000000000000000000000000000000..a10e46f94822e92088a79657ba19aa5a6eb4312e
--- /dev/null
+++ b/solutions/neighborhood_attention/natten_blackwell_h8_3x5x5.json
@@ -0,0 +1,8 @@
+{
+  "name": "natten_blackwell_h8_3x5x5",
+  "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5",
+  "author": "inference-kernels",
+  "description": "NATTEN 0.21.7 Blackwell fused neighborhood attention with inward-shifted boundary windows. Queries already include the model scale. Upstream: https://github.com/SHI-Labs/NATTEN/tree/39375ea1d2f3b1bb284315fa5f79f6ecbebb6382 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "natten==0.21.7"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from natten.functional import na3d\n\ndef run(Q, K, V):\n    return na3d(Q, K, V, kernel_size=(3, 5, 5), scale=1.0, backend='blackwell-fna')\n"}]
+}
diff --git a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_d64.json b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_d64.json
deleted file mode 100644
index 8f8374a22427840f84516ad1dc8495ab89fa779b..0000000000000000000000000000000000000000
--- a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_d64.json
+++ /dev/null
@@ -1,23 +0,0 @@
-{
-  "name": "sglang_neighborhood_attention_3d_d64",
-  "definition": "neighborhood_attention_3d_d64",
-  "author": "kda",
-  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B300"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for neighborhood_attention_3d_d64, every LTX-2.5 decoder stage.\n\nSGLang diffusion runtime at commit 684fea8a,\npython/sglang/multimodal_gen/runtime/models/decoders/ltx_2_5_diffusion_decoder.py:\nLTX2VideoVaeNeighborhoodAttention.forward (:425) after project_qkv (:435).\nNATTEN is absent from the image, so _na3d() (:49) is None and forward takes the\nFlexAttention branch: Q/K/V viewed as (B, heads, S, head_dim) (:449-459), the\nblock mask from build_block_mask (:413) -> _neighborhood_block_mask(T, H, W,\nkernel_size, device) (:170), and _flex_attention_fn() (:104, torch.compile of\nflex_attention, dynamic=False) called with scale=1.0 (:465-467).\n\nAdapter: Q/K/V arrive as project_qkv returns them, [B, T, Y, X, H, D].\nkernel_size is the stage's (Kt, Ky, Kx), which the definition passes as inputs. The decoder builds the block mask\nonce per stage (:771, :785, :835) and passes it to every block, so prepare\nbuilds it and launch runs only the attention. The output keeps the heads\nsplit: :468-470 merges them into channels for to_out, which is not part of\nthis op; that reshape is a view of the flex output, so launch detaches it\n(no copy) to return a tensor that owns its storage.\n\"\"\"\n\nimport torch\nfrom sglang.multimodal_gen.runtime.models.decoders import (\n    ltx_2_5_diffusion_decoder as ltx,\n)\n\n# The engine meets a few grids per process; one judge process meets every\n# workload's shape. Past dynamo's default limit of 8 recompiles flex_attention\n# would run uncompiled, materializing the dense S x S scores.\ntorch._dynamo.config.recompile_limit = 1024\ntorch._dynamo.config.accumulated_recompile_limit = 4096\n# The judge rejects a candidate that leaves child processes running, and\n# inductor's parallel compile pool is one; compile in-process instead.\ntorch._inductor.config.compile_threads = 1\n\n\ndef prepare(Q, K, V, Kt, Ky, Kx):\n    batch_size, num_frames, height, width, heads, head_dim = Q.shape\n    seq_len = num_frames * height * width\n    # :451-459\n    query = Q.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    key = K.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    value = V.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    # build_block_mask, :420-423\n    block_mask = ltx._neighborhood_block_mask(\n        num_frames, height, width, (int(Kt), int(Ky), int(Kx)), Q.device\n    )\n    flex_attention = ltx._flex_attention_fn()\n\n    def launch():\n        # :465-467\n        hidden_states = flex_attention(\n            query, key, value, block_mask=block_mask, scale=1.0\n        )\n        # :468-470, heads kept split. The reshape is a view of flex_attention's\n        # own exact-sized output; detach() hands it over as a tensor of its own.\n        return hidden_states.transpose(1, 2).reshape(Q.shape).detach()\n\n    return launch\n\n\ndef run(Q, K, V, Kt, Ky, Kx):\n    return prepare(Q, K, V, Kt, Ky, Kx)()\n"
-    }
-  ]
-}
diff --git a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5ddea11002b091191e3e6d2f7e7afab813ccc36
--- /dev/null
+++ b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_neighborhood_attention_3d_h16_d64_kt3_ky7_kx7",
+  "definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.\n\nSGLang diffusion runtime at commit 684fea8a,\npython/sglang/multimodal_gen/runtime/models/decoders/ltx_2_5_diffusion_decoder.py:\nLTX2VideoVaeNeighborhoodAttention.forward (:425) after project_qkv (:435).\nNATTEN is absent from the image, so _na3d() (:49) is None and forward takes the\nFlexAttention branch: Q/K/V viewed as (B, heads, S, head_dim) (:449-459), the\nblock mask from build_block_mask (:413) -> _neighborhood_block_mask(T, H, W,\nkernel_size, device) (:170), and _flex_attention_fn() (:104, torch.compile of\nflex_attention, dynamic=False) called with scale=1.0 (:465-467).\n\nAdapter: Q/K/V arrive as project_qkv returns them, [B, T, Y, X, H, D].\nkernel_size is the module's (Kt, Ky, Kx). The decoder builds the block mask\nonce per stage (:771, :785, :835) and passes it to every block, so prepare\nbuilds it and launch runs only the attention. The output keeps the heads\nsplit: :468-470 merges them into channels for to_out, which is not part of\nthis op; that reshape is a view of the flex output, so launch detaches it\n(no copy) to return a tensor that owns its storage.\n\"\"\"\n\nimport torch\nfrom sglang.multimodal_gen.runtime.models.decoders import (\n    ltx_2_5_diffusion_decoder as ltx,\n)\n\nKERNEL_SIZE = (3, 7, 7)\n\n# The engine meets a few grids per process; one judge process meets every\n# workload's shape. Past dynamo's default limit of 8 recompiles flex_attention\n# would run uncompiled, materializing the dense S x S scores.\ntorch._dynamo.config.recompile_limit = 1024\ntorch._dynamo.config.accumulated_recompile_limit = 4096\n# The judge rejects a candidate that leaves child processes running, and\n# inductor's parallel compile pool is one; compile in-process instead.\ntorch._inductor.config.compile_threads = 1\n\n\ndef prepare(Q, K, V):\n    batch_size, num_frames, height, width, heads, head_dim = Q.shape\n    seq_len = num_frames * height * width\n    # :451-459\n    query = Q.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    key = K.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    value = V.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    # build_block_mask, :420-423\n    block_mask = ltx._neighborhood_block_mask(\n        num_frames, height, width, KERNEL_SIZE, Q.device\n    )\n    flex_attention = ltx._flex_attention_fn()\n\n    def launch():\n        # :465-467\n        hidden_states = flex_attention(\n            query, key, value, block_mask=block_mask, scale=1.0\n        )\n        # :468-470, heads kept split. The reshape is a view of flex_attention's\n        # own exact-sized output; detach() hands it over as a tensor of its own.\n        return hidden_states.transpose(1, 2).reshape(Q.shape).detach()\n\n    return launch\n\n\ndef run(Q, K, V):\n    return prepare(Q, K, V)()\n"
+    }
+  ]
+}
diff --git a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b9470880036db80584989a5858078cac391d768
--- /dev/null
+++ b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_neighborhood_attention_3d_h32_d64_kt3_ky7_kx7",
+  "definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.\n\nSGLang diffusion runtime at commit 684fea8a,\npython/sglang/multimodal_gen/runtime/models/decoders/ltx_2_5_diffusion_decoder.py:\nLTX2VideoVaeNeighborhoodAttention.forward (:425) after project_qkv (:435).\nNATTEN is absent from the image, so _na3d() (:49) is None and forward takes the\nFlexAttention branch: Q/K/V viewed as (B, heads, S, head_dim) (:449-459), the\nblock mask from build_block_mask (:413) -> _neighborhood_block_mask(T, H, W,\nkernel_size, device) (:170), and _flex_attention_fn() (:104, torch.compile of\nflex_attention, dynamic=False) called with scale=1.0 (:465-467).\n\nAdapter: Q/K/V arrive as project_qkv returns them, [B, T, Y, X, H, D].\nkernel_size is the module's (Kt, Ky, Kx). The decoder builds the block mask\nonce per stage (:771, :785, :835) and passes it to every block, so prepare\nbuilds it and launch runs only the attention. The output keeps the heads\nsplit: :468-470 merges them into channels for to_out, which is not part of\nthis op; that reshape is a view of the flex output, so launch detaches it\n(no copy) to return a tensor that owns its storage.\n\"\"\"\n\nimport torch\nfrom sglang.multimodal_gen.runtime.models.decoders import (\n    ltx_2_5_diffusion_decoder as ltx,\n)\n\nKERNEL_SIZE = (3, 7, 7)\n\n# The engine meets a few grids per process; one judge process meets every\n# workload's shape. Past dynamo's default limit of 8 recompiles flex_attention\n# would run uncompiled, materializing the dense S x S scores.\ntorch._dynamo.config.recompile_limit = 1024\ntorch._dynamo.config.accumulated_recompile_limit = 4096\n# The judge rejects a candidate that leaves child processes running, and\n# inductor's parallel compile pool is one; compile in-process instead.\ntorch._inductor.config.compile_threads = 1\n\n\ndef prepare(Q, K, V):\n    batch_size, num_frames, height, width, heads, head_dim = Q.shape\n    seq_len = num_frames * height * width\n    # :451-459\n    query = Q.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    key = K.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    value = V.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    # build_block_mask, :420-423\n    block_mask = ltx._neighborhood_block_mask(\n        num_frames, height, width, KERNEL_SIZE, Q.device\n    )\n    flex_attention = ltx._flex_attention_fn()\n\n    def launch():\n        # :465-467\n        hidden_states = flex_attention(\n            query, key, value, block_mask=block_mask, scale=1.0\n        )\n        # :468-470, heads kept split. The reshape is a view of flex_attention's\n        # own exact-sized output; detach() hands it over as a tensor of its own.\n        return hidden_states.transpose(1, 2).reshape(Q.shape).detach()\n\n    return launch\n\n\ndef run(Q, K, V):\n    return prepare(Q, K, V)()\n"
+    }
+  ]
+}
diff --git a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9f99086b9e878fd32b8d305fbce77d23b915e90
--- /dev/null
+++ b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_neighborhood_attention_3d_h4_d64_kt11_ky11_kx11",
+  "definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.\n\nSGLang diffusion runtime at commit 684fea8a,\npython/sglang/multimodal_gen/runtime/models/decoders/ltx_2_5_diffusion_decoder.py:\nLTX2VideoVaeNeighborhoodAttention.forward (:425) after project_qkv (:435).\nNATTEN is absent from the image, so _na3d() (:49) is None and forward takes the\nFlexAttention branch: Q/K/V viewed as (B, heads, S, head_dim) (:449-459), the\nblock mask from build_block_mask (:413) -> _neighborhood_block_mask(T, H, W,\nkernel_size, device) (:170), and _flex_attention_fn() (:104, torch.compile of\nflex_attention, dynamic=False) called with scale=1.0 (:465-467).\n\nAdapter: Q/K/V arrive as project_qkv returns them, [B, T, Y, X, H, D].\nkernel_size is the module's (Kt, Ky, Kx). The decoder builds the block mask\nonce per stage (:771, :785, :835) and passes it to every block, so prepare\nbuilds it and launch runs only the attention. The output keeps the heads\nsplit: :468-470 merges them into channels for to_out, which is not part of\nthis op; that reshape is a view of the flex output, so launch detaches it\n(no copy) to return a tensor that owns its storage.\n\"\"\"\n\nimport torch\nfrom sglang.multimodal_gen.runtime.models.decoders import (\n    ltx_2_5_diffusion_decoder as ltx,\n)\n\nKERNEL_SIZE = (11, 11, 11)\n\n# The engine meets a few grids per process; one judge process meets every\n# workload's shape. Past dynamo's default limit of 8 recompiles flex_attention\n# would run uncompiled, materializing the dense S x S scores.\ntorch._dynamo.config.recompile_limit = 1024\ntorch._dynamo.config.accumulated_recompile_limit = 4096\n# The judge rejects a candidate that leaves child processes running, and\n# inductor's parallel compile pool is one; compile in-process instead.\ntorch._inductor.config.compile_threads = 1\n\n\ndef prepare(Q, K, V):\n    batch_size, num_frames, height, width, heads, head_dim = Q.shape\n    seq_len = num_frames * height * width\n    # :451-459\n    query = Q.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    key = K.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    value = V.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    # build_block_mask, :420-423\n    block_mask = ltx._neighborhood_block_mask(\n        num_frames, height, width, KERNEL_SIZE, Q.device\n    )\n    flex_attention = ltx._flex_attention_fn()\n\n    def launch():\n        # :465-467\n        hidden_states = flex_attention(\n            query, key, value, block_mask=block_mask, scale=1.0\n        )\n        # :468-470, heads kept split. The reshape is a view of flex_attention's\n        # own exact-sized output; detach() hands it over as a tensor of its own.\n        return hidden_states.transpose(1, 2).reshape(Q.shape).detach()\n\n    return launch\n\n\ndef run(Q, K, V):\n    return prepare(Q, K, V)()\n"
+    }
+  ]
+}
diff --git a/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9434afd793a8666c8d97c704660419033b51f8c
--- /dev/null
+++ b/solutions/neighborhood_attention/sglang_neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_neighborhood_attention_3d_h8_d64_kt3_ky5_kx5",
+  "definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.\n\nSGLang diffusion runtime at commit 684fea8a,\npython/sglang/multimodal_gen/runtime/models/decoders/ltx_2_5_diffusion_decoder.py:\nLTX2VideoVaeNeighborhoodAttention.forward (:425) after project_qkv (:435).\nNATTEN is absent from the image, so _na3d() (:49) is None and forward takes the\nFlexAttention branch: Q/K/V viewed as (B, heads, S, head_dim) (:449-459), the\nblock mask from build_block_mask (:413) -> _neighborhood_block_mask(T, H, W,\nkernel_size, device) (:170), and _flex_attention_fn() (:104, torch.compile of\nflex_attention, dynamic=False) called with scale=1.0 (:465-467).\n\nAdapter: Q/K/V arrive as project_qkv returns them, [B, T, Y, X, H, D].\nkernel_size is the module's (Kt, Ky, Kx). The decoder builds the block mask\nonce per stage (:771, :785, :835) and passes it to every block, so prepare\nbuilds it and launch runs only the attention. The output keeps the heads\nsplit: :468-470 merges them into channels for to_out, which is not part of\nthis op; that reshape is a view of the flex output, so launch detaches it\n(no copy) to return a tensor that owns its storage.\n\"\"\"\n\nimport torch\nfrom sglang.multimodal_gen.runtime.models.decoders import (\n    ltx_2_5_diffusion_decoder as ltx,\n)\n\nKERNEL_SIZE = (3, 5, 5)\n\n# The engine meets a few grids per process; one judge process meets every\n# workload's shape. Past dynamo's default limit of 8 recompiles flex_attention\n# would run uncompiled, materializing the dense S x S scores.\ntorch._dynamo.config.recompile_limit = 1024\ntorch._dynamo.config.accumulated_recompile_limit = 4096\n# The judge rejects a candidate that leaves child processes running, and\n# inductor's parallel compile pool is one; compile in-process instead.\ntorch._inductor.config.compile_threads = 1\n\n\ndef prepare(Q, K, V):\n    batch_size, num_frames, height, width, heads, head_dim = Q.shape\n    seq_len = num_frames * height * width\n    # :451-459\n    query = Q.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    key = K.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    value = V.reshape(batch_size, seq_len, heads, head_dim).transpose(1, 2)\n    # build_block_mask, :420-423\n    block_mask = ltx._neighborhood_block_mask(\n        num_frames, height, width, KERNEL_SIZE, Q.device\n    )\n    flex_attention = ltx._flex_attention_fn()\n\n    def launch():\n        # :465-467\n        hidden_states = flex_attention(\n            query, key, value, block_mask=block_mask, scale=1.0\n        )\n        # :468-470, heads kept split. The reshape is a view of flex_attention's\n        # own exact-sized output; detach() hands it over as a tensor of its own.\n        return hidden_states.transpose(1, 2).reshape(Q.shape).detach()\n\n    return launch\n\n\ndef run(Q, K, V):\n    return prepare(Q, K, V)()\n"
+    }
+  ]
+}
diff --git a/solutions/sparse_mla/flashmla_prefill.json b/solutions/sparse_mla/flashmla_prefill.json
deleted file mode 100644
index c22f87df8f05fd53227fa1c6c8db425c9cc9baca..0000000000000000000000000000000000000000
--- a/solutions/sparse_mla/flashmla_prefill.json
+++ /dev/null
@@ -1,24 +0,0 @@
-{
-  "name": "flashmla_prefill",
-  "definition": "sparse_mla_prefill_h64_d512",
-  "author": "inference-kernels",
-  "description": "FlashMLA Blackwell BF16 sparse prefill with 128 or 640 index slots, per-query sparse lengths and attention sinks. Merged by KDA from the per-shape solutions when the per-shape definitions were merged. Upstream: https://github.com/deepseek-ai/FlashMLA/tree/0f11e178da1f67becb087a324b400854931ec3e3 (MIT).",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B200"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "torch>=2.11",
-      "flash-mla @ git+https://github.com/deepseek-ai/FlashMLA.git@0f11e178da1f67becb087a324b400854931ec3e3"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "from flash_mla import flash_mla_sparse_fwd\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return flash_mla_sparse_fwd(\n        Q, KV, indices, sm_scale, d_v=512,\n        attn_sink=attn_sink, topk_length=topk_length,\n    )[0]\n"
-    }
-  ]
-}
diff --git a/solutions/sparse_mla/flashmla_prefill_k128.json b/solutions/sparse_mla/flashmla_prefill_k128.json
new file mode 100644
index 0000000000000000000000000000000000000000..a443b0fd8f001c3dbe6bb6eb9d5719bfeaed7bfc
--- /dev/null
+++ b/solutions/sparse_mla/flashmla_prefill_k128.json
@@ -0,0 +1,8 @@
+{
+  "name": "flashmla_prefill_k128",
+  "definition": "sparse_mla_prefill_h64_d512_k128",
+  "author": "inference-kernels",
+  "description": "FlashMLA Blackwell BF16 sparse prefill with 128 index slots, per-query sparse lengths and attention sinks. Upstream: https://github.com/deepseek-ai/FlashMLA/tree/0f11e178da1f67becb087a324b400854931ec3e3 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "flash-mla @ git+https://github.com/deepseek-ai/FlashMLA.git@0f11e178da1f67becb087a324b400854931ec3e3"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from flash_mla import flash_mla_sparse_fwd\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return flash_mla_sparse_fwd(\n        Q, KV, indices, sm_scale, d_v=512,\n        attn_sink=attn_sink, topk_length=topk_length,\n    )[0]\n"}]
+}
diff --git a/solutions/sparse_mla/flashmla_prefill_k640.json b/solutions/sparse_mla/flashmla_prefill_k640.json
new file mode 100644
index 0000000000000000000000000000000000000000..cabcb2ae2d2554b06a004d80ab804d6444abcd94
--- /dev/null
+++ b/solutions/sparse_mla/flashmla_prefill_k640.json
@@ -0,0 +1,8 @@
+{
+  "name": "flashmla_prefill_k640",
+  "definition": "sparse_mla_prefill_h64_d512_k640",
+  "author": "inference-kernels",
+  "description": "FlashMLA Blackwell BF16 sparse prefill with 640 index slots, per-query sparse lengths and attention sinks. Upstream: https://github.com/deepseek-ai/FlashMLA/tree/0f11e178da1f67becb087a324b400854931ec3e3 (MIT).",
+  "spec": {"language": "python", "target_hardware": ["NVIDIA_B200"], "entry_point": "main.py::run", "dependencies": ["torch>=2.11", "flash-mla @ git+https://github.com/deepseek-ai/FlashMLA.git@0f11e178da1f67becb087a324b400854931ec3e3"], "destination_passing_style": false},
+  "sources": [{"path": "main.py", "content": "from flash_mla import flash_mla_sparse_fwd\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return flash_mla_sparse_fwd(\n        Q, KV, indices, sm_scale, d_v=512,\n        attn_sink=attn_sink, topk_length=topk_length,\n    )[0]\n"}]
+}
diff --git a/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512.json b/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512.json
deleted file mode 100644
index f99a304c07cc86dd0110f79255914988828aff36..0000000000000000000000000000000000000000
--- a/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512.json
+++ /dev/null
@@ -1,23 +0,0 @@
-{
-  "name": "sglang_sparse_mla_prefill_h64_d512",
-  "definition": "sparse_mla_prefill_h64_d512",
-  "author": "kda",
-  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
-  "spec": {
-    "language": "python",
-    "target_hardware": [
-      "NVIDIA_B300"
-    ],
-    "entry_point": "main.py::run",
-    "dependencies": [
-      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
-    ],
-    "destination_passing_style": false
-  },
-  "sources": [
-    {
-      "path": "main.py",
-      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for sparse_mla_prefill_h64_d512 (TopK 128 or 640).\n\nSGLang 684fea8a, DeepSeek-V4.1 sparse prefill with ``--tp-size 4``:\n``DeepseekV4AttnBackend._forward_prefill_sparse``\n(sglang/srt/layers/attention/deepseek_v4_backend.py:3848) calls\n``sgl_kernel.flash_mla.flash_mla_sparse_fwd(q, kv, indices, sm_scale, d_v,\nattn_sink, topk_length)`` (:3916) on the BF16 workspace. Each rank passes its\n16 query heads in a zero-filled 64-head buffer (models/deepseek_v4.py:2342,\nwidth :1054) with ``attn_sink`` sliced to the rank and zero-padded the same way\n(:1056-1079), and keeps its first 16 output heads (:2492).\n\nAdapter: the four ranks run one after another and their heads are\nconcatenated in rank order, as the capture gathers them.\n\"\"\"\n\nimport torch\nfrom sgl_kernel.flash_mla import flash_mla_sparse_fwd\n\nTP = 4  # capture: --tp-size 4\n\n\ndef prepare(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    heads = Q.shape[1] // TP\n    q_padded = torch.zeros_like(Q)\n    sink = torch.zeros_like(attn_sink)\n\n    def launch():\n        out = []\n        for rank in range(TP):\n            local = slice(rank * heads, (rank + 1) * heads)\n            q_padded[:, :heads] = Q[:, local]\n            sink[:heads] = attn_sink[local]\n            o, _, _ = flash_mla_sparse_fwd(\n                q=q_padded,\n                kv=KV,\n                indices=indices,\n                sm_scale=float(sm_scale),\n                d_v=Q.shape[-1],\n                attn_sink=sink,\n                topk_length=topk_length,\n            )\n            out.append(o[:, :heads])\n        return torch.cat(out, 1)\n\n    return launch\n\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return prepare(Q, KV, indices, topk_length, attn_sink, sm_scale)()\n"
-    }
-  ]
-}
diff --git a/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k128.json b/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k128.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dfa3aa17314a60403a7a174afed7ed1b2bef461
--- /dev/null
+++ b/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k128.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_sparse_mla_prefill_h64_d512_k128",
+  "definition": "sparse_mla_prefill_h64_d512_k128",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for sparse_mla_prefill_h64_d512_k128.\n\nSGLang 684fea8a, DeepSeek-V4.1 sparse prefill with ``--tp-size 4``:\n``DeepseekV4AttnBackend._forward_prefill_sparse``\n(sglang/srt/layers/attention/deepseek_v4_backend.py:3848) calls\n``sgl_kernel.flash_mla.flash_mla_sparse_fwd(q, kv, indices, sm_scale, d_v,\nattn_sink, topk_length)`` (:3916) on the BF16 workspace. Each rank passes its\n16 query heads in a zero-filled 64-head buffer (models/deepseek_v4.py:2342,\nwidth :1054) with ``attn_sink`` sliced to the rank and zero-padded the same way\n(:1056-1079), and keeps its first 16 output heads (:2492).\n\nAdapter: the four ranks run one after another and their heads are\nconcatenated in rank order, as the capture gathers them.\n\"\"\"\n\nimport torch\nfrom sgl_kernel.flash_mla import flash_mla_sparse_fwd\n\nTP = 4  # capture: --tp-size 4\n\n\ndef prepare(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    heads = Q.shape[1] // TP\n    q_padded = torch.zeros_like(Q)\n    sink = torch.zeros_like(attn_sink)\n\n    def launch():\n        out = []\n        for rank in range(TP):\n            local = slice(rank * heads, (rank + 1) * heads)\n            q_padded[:, :heads] = Q[:, local]\n            sink[:heads] = attn_sink[local]\n            o, _, _ = flash_mla_sparse_fwd(\n                q=q_padded,\n                kv=KV,\n                indices=indices,\n                sm_scale=float(sm_scale),\n                d_v=Q.shape[-1],\n                attn_sink=sink,\n                topk_length=topk_length,\n            )\n            out.append(o[:, :heads])\n        return torch.cat(out, 1)\n\n    return launch\n\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return prepare(Q, KV, indices, topk_length, attn_sink, sm_scale)()\n"
+    }
+  ]
+}
diff --git a/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k640.json b/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k640.json
new file mode 100644
index 0000000000000000000000000000000000000000..d970832ddca9352e67c3d2286e84a1db00cdb3f5
--- /dev/null
+++ b/solutions/sparse_mla/sglang_sparse_mla_prefill_h64_d512_k640.json
@@ -0,0 +1,23 @@
+{
+  "name": "sglang_sparse_mla_prefill_h64_d512_k640",
+  "definition": "sparse_mla_prefill_h64_d512_k640",
+  "author": "kda",
+  "description": "SGLang's own implementation at the capture commit 684fea8a, called the way the engine calls it (see the module docstring). Run it in lmsysorg/sglang@sha256:ddae0e399d3d94f7827f16183301cf8dac4cf0eae0ef25590b7dac5881d713d8. KDA (Kernel Design Agents) times it live as the baseline: prepare(*inputs) does the host-side setup and returns a zero-argument launch, and only launch is timed; run(*inputs) does both.",
+  "spec": {
+    "language": "python",
+    "target_hardware": [
+      "NVIDIA_B300"
+    ],
+    "entry_point": "main.py::run",
+    "dependencies": [
+      "sglang @ git+https://github.com/sgl-project/sglang.git@684fea8af3b11a5b51ec9966c02959a32ad3421c#subdirectory=python"
+    ],
+    "destination_passing_style": false
+  },
+  "sources": [
+    {
+      "path": "main.py",
+      "content": "# Copyright 2026 Co-developed by KDA (Kernel Design Agents)\n\"\"\"SGLang baseline for sparse_mla_prefill_h64_d512_k640.\n\nSGLang 684fea8a, DeepSeek-V4.1 sparse prefill with ``--tp-size 4``:\n``DeepseekV4AttnBackend._forward_prefill_sparse``\n(sglang/srt/layers/attention/deepseek_v4_backend.py:3848) calls\n``sgl_kernel.flash_mla.flash_mla_sparse_fwd(q, kv, indices, sm_scale, d_v,\nattn_sink, topk_length)`` (:3916) on the BF16 workspace. Each rank passes its\n16 query heads in a zero-filled 64-head buffer (models/deepseek_v4.py:2342,\nwidth :1054) with ``attn_sink`` sliced to the rank and zero-padded the same way\n(:1056-1079), and keeps its first 16 output heads (:2492).\n\nAdapter: the four ranks run one after another and their heads are\nconcatenated in rank order, as the capture gathers them.\n\"\"\"\n\nimport torch\nfrom sgl_kernel.flash_mla import flash_mla_sparse_fwd\n\nTP = 4  # capture: --tp-size 4\n\n\ndef prepare(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    heads = Q.shape[1] // TP\n    q_padded = torch.zeros_like(Q)\n    sink = torch.zeros_like(attn_sink)\n\n    def launch():\n        out = []\n        for rank in range(TP):\n            local = slice(rank * heads, (rank + 1) * heads)\n            q_padded[:, :heads] = Q[:, local]\n            sink[:heads] = attn_sink[local]\n            o, _, _ = flash_mla_sparse_fwd(\n                q=q_padded,\n                kv=KV,\n                indices=indices,\n                sm_scale=float(sm_scale),\n                d_v=Q.shape[-1],\n                attn_sink=sink,\n                topk_length=topk_length,\n            )\n            out.append(o[:, :heads])\n        return torch.cat(out, 1)\n\n    return launch\n\n\ndef run(Q, KV, indices, topk_length, attn_sink, sm_scale):\n    return prepare(Q, KV, indices, topk_length, attn_sink, sm_scale)()\n"
+    }
+  ]
+}
diff --git a/workloads/dense_attention/noncausal_dense_attention_h32.jsonl b/workloads/dense_attention/noncausal_dense_attention_h32.jsonl
deleted file mode 100644
index a3a56902f2a99f70534afd256e793cd2fba64ccf..0000000000000000000000000000000000000000
--- a/workloads/dense_attention/noncausal_dense_attention_h32.jsonl
+++ /dev/null
@@ -1,162 +0,0 @@
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c884d64e-9420-501c-b198-7807cdda2443"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cacf8279-51df-5a97-a592-9d328a81df13"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2a358cbf-1566-510c-b15e-274202339d18"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8471e8e5-d50c-51c2-b946-b92561322441"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "475965cf-cbad-5baa-b685-479860697719"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4795bcdd-6f37-5076-9404-a44be707081c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "abb09f82-ee47-528a-905a-f86b658621f5"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "645dc728-94d6-5801-a837-9a1f4201b237"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 64, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "10493d50-a0f8-527f-8d39-414c377360f2"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "a34594b2-2c8d-58cd-b443-630151fec624"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 1, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 2, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d3373710-c368-59f3-9515-7b961b87d57c"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "919246a0-b394-5215-b9c0-afda319ecefa"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32"}}
-{"definition": "noncausal_dense_attention_h32", "workload": {"axes": {"B": 4, "D": 128, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f"}}
diff --git a/workloads/dense_attention/noncausal_dense_attention_h32_d128.jsonl b/workloads/dense_attention/noncausal_dense_attention_h32_d128.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..f04c34faf4bb792021fee725f1dec8c04b0d022d
--- /dev/null
+++ b/workloads/dense_attention/noncausal_dense_attention_h32_d128.jsonl
@@ -0,0 +1,54 @@
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/2194b3ea3a2317a736f4819d874cf30d964a71b814ef0137f92ba9094eabdb1b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/46a65d3099b0c3970bea5c9054193b11ab7401eb7d9dd7cdb4472691a4986226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c8a8ba441616aad961e35f3aabebed22bdf767022825b9ed18f17d081a452e9d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "c7ba0aad-2212-5bb2-8896-64d663665b59"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3cda2cdcf1e4a37601f0354689cc38cff44660eeeb1d84c380e6bf44b4698bbc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/03fc3b844867f6e93c14ca78983b2f2f9ec71629f603d31537105fb53a9de36e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f65243f89275b0e61058e62ed7a332a931ddcdf016193d50b55415fb66e68eb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "40268ed4-06c5-5827-81ef-1c9ee0ee1538"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/fdfb909b080ede98a45c0af0c49a079ad9198f35cb83651c5bc3f1e2467fbbf5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/88d3655c684df025dd7823274bb500459a998b241cbf58077ddbfbf139b03e4d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/8e7efa54ba5a1a80f2c6051826649358c9ff133d6611712bc130a2bf05c41cce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "cb8dd366-a40b-5ea5-a035-80beb2c63a25"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3c0a6e88a222e246fc0838ad4b7e27cb1132a960921c1b43b77245bf0d88dedb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/bca1a91fb5c61e44eb60e2e2e81aa7fa5dde7a6375a1be591af588c89e213ce8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/093192d3cb551e2e11c5cee0b17978c745e5212d77b19bff98af3f4333e42449.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "c80509df-cd6d-52cb-b405-6c4eb5efee7f"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/61f350ac17f26db3310d1cf68502c937bed6b5228e9335189331b4c7916f398d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/878c27b4d62f03a04ab7ae7b7972b98f9f3bb687efbc6404f08a5725ce5d5792.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c503c6e9979a1719c5fd5fb8b278f36c46a7dcbc2123e7a8c90200016a09197f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "82872221-aef0-5c8a-83c1-8c116c94a80b"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/187041444f4c15f301c11c49afe657e6bb9d79942db6ae3947ce02582302e57d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/71aa524fa5d7eede111cd694d35c47874783a77548c5f525c2ab755ef22479b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3ead71b413d3298ce398f2d15b91de9225fa3d47b502561d79f0e863cf49e0b5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f849f163-b4b3-5616-84cd-83f86c0d1187"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/eaee199697221a714711b6160a435d0914f26445c8e47779da939f88c00328ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f55b765141b56ff4c53cd619ac30f13d4671ce4fa0e4576823f99de47b1d5adb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/d281ea4f5cf9e7afce398582e40fced8927793e546f56e280c4981770af0697d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b76a8861-ae59-5f3b-92b4-c02019ca4941"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/001a38fd35aae69255cab17e2549147c9976ab2a8bdc49e6aefc2d354995dd58.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/44902634cdf02979d1888dc498aecd8515ae009a479b01dd2ad76b04ecc8e5e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6d544b9d5ca28d910d69b67e9e2e19be7147fb9595bbdd22a1f33c286545a72e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "43bb2e12-3ee6-5e23-9192-040a8405a7ae"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b939399b03c23b9bcb40649e0d57efd578fc83413662949343370e780622e339.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f0537d7965af8e06683158c8f64d33c7cf7546aa160f29c6ca956bc2812203dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/aa5808be14012da240a10e27a858bd49a22b523b4ff86d2293f551a18b4967ca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "68caa03d-dfc2-5247-bee5-37a69aa2a52d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/08a9f963eb0ffab82adcc61c302ce1f3299453dcbe39da2b71a8462a863a7565.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9c39ddf9ee9ceebf6a8d1a7b7b6f7dc134fc813101d6b09be28a593cef724eed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/521ffb15d0b3a543cc52dc8dca21b4c5f6d58b8ec5c06d169a7e4fc7721f73b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "10493d50-a0f8-527f-8d39-414c377360f2"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ced2ccd7e8038edb6d4502006274e231bf49f46e4f161a11930dab19b263e3e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5b887af498c1beee282fb33161274a2b70b42abab4095ad99c4d76dd776a95c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/64fff20d686e5a575527b5e8a2d01ccfb749552b7780e2e5a4642bd3ed78ea7c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "328c3d7b-76a2-5bf9-a9be-cedd6c2c58d8"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/191bbcdb482b55518307aa2c3cffd362f50118d5446ba49ee869c34e426b401c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6bb522fbc871a5ae13083f5b3fffd37711c32ad63ce11f511fabb8094c357c7f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b36f5f4d6bc8ded513a2bf7aad7dda91010bb455a9be3bb81e924f29d87b2876.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "ca1ac853-8f3e-5179-902b-06ef7d561703"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a018391e05d448d7a1b119f3cdce804182646d3143f480d07f317b0f837d3e8e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6fad84c78e0276602b9c0f80451bc3a9698cba3c1bcb67e4505b57f204ac70d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/bd7bb79e6651d1b92d6480e3ca519fd7f6491217511258cea8cadd711c2e8f37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "22d5bea1-549f-5d31-bb56-4be38b775291"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7a1707397e8a9e789983d69fce177e7ecc4311d1da72eb083c66c59f737f6ed9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6c61ceedc13fac1000104efedd4859c127d0114c2a824644e79e7663fd07979f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/db708edf04be27d0fb319b1bd461515093da5333e15eaac5df1249237d869064.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "93a38d68-1ec7-54cf-9bf8-0d46651c3adf"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/388a7bbb174c8af340e8887900854867d0e85183f071ad77f2a6efa0ad8e7be6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c793c49e85903c919d5760613015ce5d75c9cb0a3f81dbaa3ef0db7dbf93ed06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a698701caf5aceac8ac39b6a0fb6d3c7368f6fd4d727b0843082cd34e38df705.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "a34594b2-2c8d-58cd-b443-630151fec624"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5f320947d02423f681e47151c19e09a72b53a31a68f6e9000cf4990b8e22c2d5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/265e075117058320933586f4555b3d27a81823b2e8b710cec12e0fca01670824.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/61559e1f0a0be96cdffe0347baa767f01699aa7829175735af7c59e6d5da9e51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "a5c07816-ebf1-5d84-b114-768db1f7f14a"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/8922b688f992600e9a21022b450ef79fed60ad79bc6bd7dea2f594938dcc5060.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/622023ba926fdc0333d8cba57322d3e76310bc77dace8bc6c7ca963ade2c4ce6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/624e11a027e745d3584e63be583dd655c0a8ae2f56762f05a8e75942769d1b64.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "325e1a4d-dd1b-5634-af96-f4344393a4b3"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7727b7b463b65ba912876760f5bc3f6ff9459d1372443473fb64df1a3a052d91.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/70fa263450feeb10162012828069cc75164b2a0caaa6e53852168f6ea64d21e2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e1470bd8b512bfa9ad2975fd8af843ef019b890f0a5f05927164098e1c216849.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "27af7d20-3b72-5881-ab32-4a3ab8d2cdd0"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ee086bc58dc59887688a83c17057371adb05fc98f2d91b49f45fac4a76c7784e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/76cf1ed34b9bc3fb73bc21ab964a5d63864e0b49891f10a57eec8c1e21dfc0db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ac99af19a5a29273444c328d3b48f4c582ebdd4e74ced106196854f942b42e06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6c3637a4-7d4e-515c-9f0b-eeaf53ed75ca"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6bd7efb945b16b7e8f98f94b45f6cd9e3028d865ea3286e68ce7251979b20c46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/38e10e917e86c4148c1712effee882c87150a75945543ee2fca81e4b51e9d108.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5eb906a1eeeef59e585588e8a0e12e67ac2eeb0a724a358719b9676b8c651d48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d617c58a-373a-5b40-9e36-e7c009786df5"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/40739ccd25b80fb31c05051fac0bfd5d615446b2de6f9b733926d052d79f5e94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9dfdc0f9ad325e2d12523cc6cbe63150e5c5ba117a98b85a7c9f4fe6f0c5e2fa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/62ee86e7ad4262791f664f7a9a9129a22b88d9e2eec224e703557a7b3dbbec1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "9f69a6cc-1028-52c5-8516-12d5f44e17ea"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/575800697bd6fe97718bdf60b5deedf3309208f99d104da8d62c4f7fcfd6c8f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/95a806a37b5a53b725a43a62f9854a66b9a526fd99b2eff64b92767428834989.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a2f0294fae77b8572f1c27203f9948099ca9efcc59ec6c0b8f4ffc6a8c4c4265.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "06131807-accf-5812-a37b-8f9ae948e0f5"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/de6dc71fb7a0fc970e128e26b574556fd382abf60ac95a5386ff29290f363358.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ad1785f88dbc5f7adc255af98251691c1e52c907a4f9c940f2cd2851baf9d0e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/db64aebd7ed15362335a98171db5b0af1d80d3206ecf4d564ea7a540f17e89ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "1802abf8-29bb-5236-85dd-195a445b9ab2"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/11444bc151df80805277e391f13564d6efccc7c79af80c383ac83da1e0fcf661.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a3150ee97cf767d2dfe6cf93c9cbb412befcac12fc35eb7148e9f759cd523c94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/322b2d08b249bfa2d434f89b00589d2f34dc04d772cbd423f28227fa8356ff24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d7ed325e-48af-50dd-a16a-d93ebe4d1926"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/421f1d069d3590d453991524535be20375bc76fac451676dc84e4e2b7be579bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/830c64de64954208df7fe50762fa7f9abe0c0d68084764648af6e53665dda540.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/530848649b1f77692d4aa36ef92b66103942576133c25699bad6145712c9c049.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "695ccefb-838b-52c6-8d57-7e5f6a5b755f"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/527e4b1b8d1db9763ebba5173727c7cc6e035703457dcf2ae4c69a4a3a634cf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/066d711712984ec769721a2c2cf3ffd061129b896838c1da6de7563cbce95395.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/dcf393765d97b24c1ddaeb102635494fa48d4d31722c3f7fee5b41110636841c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "72c5ddc0-a738-575e-9cce-35d0c4d7430a"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/1f58a4000564db1bb258eac2c8c79b8bcf2c09baa664319586763f2be6dadcb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/dc7245baf8c1360fd62e3d98d4d0405b45479da066d0fa845715f7345bebee25.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/05e2582fa1dd44e41bcd9b7094585216a4bf337f37e98de52c653e19bd67800b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6e7f3c9e-ea8c-519d-b268-8c1b5c19a937"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/2dbd73d2ac10bb8532d34aa54cbdce1e71df237e35a363a831a5f9ef28b1aee9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3d7f72e933f062278238f9052db30403a0c7f1bfc5b0af99321a7e55d1d7546d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/996e0538b3ef45c6dc50421d8127d58c0cc40d91422c02a363547edcc2e10e45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b7b2749e-15a7-5e24-a748-28c0329e6220"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0cb5fe890cae8735b3bce9b831e267d05eacfe0917dd56d11004ac82d94cf174.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/2469abd1db3f23dbb7aae7823be78c9e0ff480f17faee6cecf7f0ca3051276f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e8c537529dcf323f7a1dea3c227e02b334f859acb12a2190948791f7cbb8c318.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "79c92abe-237a-524b-8619-6a2084a92d6d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/899afb32fd50817547d1d6afd0f8efc4e297e061bc00f20ec8a283bad47091a7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/80998481e4bdd152a714f465d2e104b7fd4c54d3bb47922f0fac86454f9afdd7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/538982cc4d96f6e702a5e7cce0bab9bc35f9451946fb00b14403d2c7f5c7c353.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d48a99f0-024c-518f-9eb2-b0fdf84a524d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/69b940393b8ca45daf6f6dad8423ea2790e624bd8ba0c3fcfa9de9e15a86a9f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/d28ab269f9c4c6146d0c173040b6d73f97a99e2499ba15a8374b1ea3d857ecef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7d7c367157355e3461c62d585b0d6f9816f09950bc85a576d225e5ffa866af50.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "195ec930-50e5-5ab2-b6af-ce9bcede4f65"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/314dff5632e3e991ec8d6d7da1786d208774d07fc0471769db78bdeddf0709a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b0aee4fa5d6c9c869d2bcb8630306746624cf515564061f52fc51d6456146256.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b76a1a166f28f83111907fc62074e58342a799f112078efacef549db113f81b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "14f218bb-57ad-5372-a8fe-18fa938c250a"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b721599a654fec0551debc7dfafa6a2d27fdcdc568b4aaf92f4501c2b80d2469.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/90686d629cdbdccf53ec58a90f7b9ea7704924db73e7246227758fa91b5ab41b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/34b90c503ef36e7fc33ebbca019c178459e3772c01bc1fda7db3d6144d60ac4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "5188525c-ae23-5e83-aaa0-7233b2d1be4d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/820a88a96ccebf9e8053dc0693afe2d5d55fa08e50ab37e9722e79b9fe3175fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/41973e98f0c44158f03da2c594b2a5266040f6899dcca1181ceb43f1e77484a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/867688a2a7aaaf2fdcad1f96102a0981e5b2e80d781e414ca4332a4738625e3d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "b63f6ceb-944d-506e-97a6-aef2e8e454d8"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/958b03d395c37cdc07cdb3f163806a8d2db71aec56da1e8f89843e6b9354cc0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0bf491dd7df8072eb0c9942042cd9f1140182af9246f93be3451435366d80764.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/1ab79c62cb764edd7e81ef932617bbfbc32c681a43f5d41ef8bb4cad47d4b9ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "e52e3dab-ac58-5101-ab27-5bdae9bc911d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3b4e5cbe1781ad8e000c8e89fc629ced25315187ca8a0eeed6e42bc29826de29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6ce14e5c2c20195a51ec57dca5a603f8f38f665bb719dc3d95987a4c29848198.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ddb17db225aa936e01290807d5ff5cf1b59862fea6ef09fbecaf4748f7b182c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "cbe34697-5e71-573c-a820-8ff5d93ccf3b"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7ed6d3ba58fb35687b309e736ff2b5a9ec62777e6bc88cbf67fb69e44b109263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3ac8dc2c616b114b8710be3a9e5e0dd8e1f9d3a84f110f800afa1b452f09b97d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/14a3af8c7bb33764b66f7e1e51433089d3ee263495da413baeca10834cc9f45b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "536d21aa-6cb4-552b-a371-cef96f27abff"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/cb428784899064e1703e244a59b75760755ac4fdb1523723cbd3abae6a611a56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ed07f2b16be1e3539444125c6fe3f5774b206610234936c982d871da24a32715.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/03913faacd88d2d26e8fbe021a1fe1ca9a6b7942f966fa17d0a8f5bee36580a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "32d4a255-e961-5f7b-bcb6-7be48f573e74"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7db9948c60237ea4d94d4f328b1d15fc6617e1c2db64d20796934150f2c34226.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/6e135da8b4adf6c8dea3ac84dee4695d3a91cfa7fea036eda83044e4b62a8a9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/429bc99c7ad740193cfacb872c3a5fd7fabaeeb278d3d9a86a08a458c021653f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "7b693a0c-422c-53af-9ac1-146ee5d09639"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/7e67068707b3f9d86bb11e884f0c3537afe922250786fc729ecf7bd09c268d43.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/480bdb878e6676ef8a938ca3a573d7fa056ce89f14000156365b8dce1223787c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/cea18401ec8e3a5181ba523eba844bc7c9c958f0e3c65b5b7e0b9ef8fbdf11c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "da56e090-dd99-53a9-84b6-0e98333bab09"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/745aeb69b99198e45bbb14a9d738e6c04a93af17fcd20208780941daa3f6777c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/1b53b813bc30bc5c13734bb344a9d37f3548f36cc3abb8b3cf228ee69a2807cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/32936f5d66d027564deb63e2ac0ec509c39ca02e5c0f04c32d2ddf138d5bf2e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "991c130c-f455-5544-8774-7a0bcb71ca17"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/69ecf0ce1ed3d5ce6b295dd95d105bf7815fd429138e359452ffd3448ad935b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/8d7d85d066e32d30b87a19bae2a240c545712805a823a78aed1ce8948de084db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0141198f247e65746f0643834d72087fe764697bc348d7ad78c8d4842def8f44.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6cb3617c-6d8b-5c31-bf25-d5220009223e"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/daed67d4ff9190cd595b29042682a2f3c810e2b15fa38ffa3724e55010c05539.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/75b8e5126ee633ed309999eb99f11a0de7b5b20df501e578552356da3b5cd65c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ed1db3a2fb361a97f4b7f9a127530c2da6744a04302e7036d5db5ead5b540c2e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "d3373710-c368-59f3-9515-7b961b87d57c"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9d875b02297b5bdb3fa5c69b44bd5ce53cb780c845b28da1458048645023e4b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b3052343e1ceb3a18c3f30cd28fe294daec3e6489ede3accd9043cebb6a84096.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/023a9ba427556068e94eaea0186963da9238adfdc45a2764997c12a232108605.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "4cdcb87c-f1a0-5604-887b-c7a07ba2e4c4"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f71a1783a78f307337b43b59468d7209622c7a7c40f509df0118274c417e3bf4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/595f18d3614b6170f6ff47b1c1bb26edbdac7c8c124d21374b5b22db1c42bd5f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3ecad76b8d32edf4a5e9706e2474f7fd69367158ec84742c651edab00bdb3f99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "3f618d6d-4678-54d9-a3ef-3b69bb773f14"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/28f70402a0665993d20632852cdfaac1a41b5dec0058043c2af41503bdb75f76.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/4f1fbec4624f64be335c0b382707477b03b280f19a9b37f92236cd6496a9452e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/a7f10e1bf0eb33899e6054c26052f906a44c46bd98806193afc555e1a73dbbb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f08aaae1-1cb4-5945-8f8f-00afb03f90a1"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/23135f1baafecafcb84af98c3edad442d2f25fe107ff469a9b323dc7508834c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0ecf374db42312f14dcc363ac6338cd8754aae43aaf8a4e2e429eb56d6e42a95.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5b160994c51663e20d30d84b1240fdfe69468ffed57f0097ad2c8dff214c2cf1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "080b6a8e-f168-5131-9f1f-a4eeaf37cbbf"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5198beffe3bd6053941529b04fbe71fe5138ce86cfb3e347ff61b9307812d11d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/f32dc0a3c0de7135e9c9c519cdca61ca12aeb8bbd92bb776e6ab8c3f7cf8d293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/fe18f50aa95acdce74397d636c1cd2789835377959d78598e91b876d96026a09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "e58ece45-84b5-50dd-95b8-2becd9cb9d3d"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ec4bac366772a0549af07f1cf9fed43b2f4718d68fd669fa6c76714a36382d82.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/9bb722ad9cd6888a2a0c2fdc1b4eaa233607fc4aec54276d8419b96885ce1a88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3e53ef10faf14927d4e4d2214b6b0fa29a10357f760ec211019c7cb9b288adb6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "33f396cc-71fe-5f37-961b-20b86a8c078e"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/b06bda15c7eedc9a2b6ea831f69c568d9356bfb7fbf58fc79d83763f573bfe64.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/e31f0e19beb727e7b45af62642dc61ca7339408fe341aab3ad5ad8d64d3a1173.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/3c756dfa542d8bad1636ebf6d484192285896a495e8b10639dde9184fe91b735.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "6ca03b08-4582-5408-8c48-6f4b30225045"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c8922e944e1b4b7d99bc85a689210ff2393016e1ce9013a2063ad609844e8485.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/0e3c7ff477dc2aa2e1db34c152164fecaaad0bf7dd54a62f34d7f156ba679681.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/50fa281a881c6ae73dea8a76c97f62c0647c3b78c166321f05d8d641e712bd68.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "919246a0-b394-5215-b9c0-afda319ecefa"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/2e420ae05246dd3c5c52f048c84be586a3d4bfd11c7f7a375494433e6143c746.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/c990422775465d82a781b16e746f886db76d56605f805bb863e01ff5b09d798a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/cb756e4901c1b8fbaecb895976939367f689b602c5982bb833e474a311784833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "f0dfee8e-8c32-5004-a09c-248d46e01062"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/ffee69963b79d2e8b6e4e7f27685e6b7e353bf5a65fe8a3538a98f661f9c7e6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/315b2d2f837bddd8eac3a301e22f8fbcf84e20b6f934944e05048e7bcb858338.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/665e2fc409d26f5e771bf8b58e9b2d1a0a87d800928f802361e70d9e118219bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "2994bf07-9d15-5981-9289-2340dfb17b32"}}
+{"definition": "noncausal_dense_attention_h32_d128", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/24584b7ce955607be0581989225573baa46319d1eb9f41a18a80322dd4e2d8f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/fe40732ebd610262d8be2146786faca95dafd029ed55eaed6f6c3887ba254324.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d128/5e8f7ef620c3e837aa5d6421c603845dc5fb65876541498ad39dae3e5e585b63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.08838834764831845}}, "uuid": "637331d6-ea8d-5f4d-8a22-d4135f99cf2f"}}
diff --git a/workloads/dense_attention/noncausal_dense_attention_h32_d64.jsonl b/workloads/dense_attention/noncausal_dense_attention_h32_d64.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..86cd1f1f105f0734be3e2d84bd52d41042f830cf
--- /dev/null
+++ b/workloads/dense_attention/noncausal_dense_attention_h32_d64.jsonl
@@ -0,0 +1,108 @@
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c884d64e-9420-501c-b198-7807cdda2443"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/207853db921f5265010180dc92c2c470252b225c9a3e02742caec331083e16b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/14808e8e37057f4439ccde924dd80113866f5f64016f0e2cb41ebad6fc460239.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f9ae066-bf34-53c8-bb42-abe64c739e10"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0e0c5ea7cd20412bde1931b7c0a3eac69bd87e903e892705ce4d3a005e554e08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/faa04467e6679dbff9eb7f52e18ccd47bd06e428f4598cbac526d9d18d33f2a6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9495819b2c77957908ea99b73b2c9db2c13980d90e7c8d086a3e303855dbe50c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bfe2c0d7-2dd2-582f-bd46-af8b3b9dc217"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b351cef24e7ad53f66356b3e21599366c5aa0256ed75fea66eab2fccb53bfc12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ccc8ca9f54f5e195d6a155e435c24269cfec20623e5f491e8d8ae9ebf4536ae2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a5dba58a118564202a7cc0930d63aac168a2468cfe32e5949b5eea6368418c69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9182253a-2949-52dd-8da2-d2ebea4c91fc"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b9af680003fd3659b6c21c568ee16cbe61064cdda0af84e36962843ded6d1065.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bb93f0a792174fae46ae81bdd6152a545c9e4bee9fe51860f315580cc45924de.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9da72d16ac4317ca302693c9b30c3c4d2c9fb00982ab2f3ccb3d78eac29c23ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "82ffc151-9721-5a67-b4a1-2fa8c886b84b"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0211591872d0d61763ccac18fe6d8f3d4640ca1b0a1661617b92f00c27ebf1d0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4ffaa07024fb1bcc772c8a4cb1980b5fab6adb99cf053d92bb4fa5d4ad051a6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c1bb8dc404cbd523fb5e3e08e0ee6734cbec7d856c2fabd4db74afdf28b2923c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d4d64ece-f3a9-5c54-8d94-5c89afa12f64"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2e0f6c6c5ac96ea26b5925c7b168b147c6f4a6db105919d06ea6296569004aed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a3c9a9a77b5c410221585fa3eec909284a29de67345fa139a847fedb09164cbf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6a634558749c0b40a7b00382aa3242446472157e8fcc4f1771c3d0f749fb7786.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "22824146-c2bc-5edb-8ec1-e14d5a6ddfe7"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5fb16fa4adcc3c6f381185e094ae445c8f2f3bcd2b987e2eba2bbb1da17be7d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ded817021bb679916115bd5a41a50bd7eb4a3dce292779936c0ca2d2c4d648e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/eabfa09ce1090f6ec1350bbfcdc77924c394d5b9eb6173d8aa04003498bff65e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e7cff504-efb7-5e51-83ca-394ac66f04fb"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ac580ae9d76593d3684aade353216d2982b0e17d7a7a2d80b34212c7497b5073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bde84cef323c6f3950e43e5c81804f2445a773803121b76a7d2b4b2074fcdd66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/90eadda8ba4ad6cbc2506b904f6f196ae3f1d7d6b42b65d18cee9fa77c3ccc38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c48b85f2-bff4-5f3c-bfc7-e78c995ee073"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2bfe61af6d432e6ebe511372552fdf7cc55a6df0f3d4351ec1c7b80e45e9d481.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d867ad05d0e812d799bc3a209d84343ebe241573853681ddf5cad2455499c6df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/20ec4bf0f4bc30927f9cbaf0a20f0db4944a8de65f6c9d44b2f1f79b1a8b7d56.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9a5a6acb-75bc-5ab6-97d2-91d7c78c2483"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7ed7e5435b505e30e6b23b2da3f5ba1d890ed7991a338abc84e405852fcdabca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/62e9d019ff507cf809f1e0fe46772de8248fda0a3377da5ea61e2fef8fb380a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0abfd5695da71e9d50f9ba9175ac46949fc37d90ddf15903819f1993e07dd70e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "5ecdfdff-3ac2-56c0-86b1-fe3a456700ee"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2daab140e440de2de7a828b1f6bbdbf5559540cecc272e0b0347c93383bfaa18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8175ab1d1f4aeb32ed17fb688aaf61bf14c630895451e1bcb6cf0fa01f599b53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6dfeaae654ec5392c1fd71a31b9c96901f0960da0ad30fe4634d13b07728d957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b62e8d3a-b28f-5b6b-a9bc-51be5ad74e5e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f6db002ee61f456a3a4d0ad4974f29ece53ddad35964552e21cac134444d396b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f8b6b54bd591ded6464899ba8d342306a022bfb02b46e5d595d6a1e549e1bb22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c8bfe5bf688af53ad27a37c46f9acabcda8757ac9fb1ff0f9622c3e56250794e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f78b3e90-16eb-548e-b877-6d9f4d4d7acd"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d184b7887eb156dab1e2329ed8ae201bcc85d154b2fa116fd6b41c841ea3e814.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a324689a488fee1512f747c5c472da874ddb27d7c389a64f17f316a66ac2b60b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e18fcb376253906b7b0192bde75dad08312bf35fa897464e0ca401025fc18fbc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c59e00ad-a2d0-59a9-9ad8-8760960df7e2"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8cab7bfd175643884b34e7a2387007b4950479c72fc53baa50fe30dc2a59176b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0163558606c89e7034507b7b21eb41c29048e9dffd849ebbb441f3e60511ca48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0b3f4496c6c682f123ceaee98bad35abc879f831c8689c15257bf3c67663971c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4522fae5-661b-5625-bad8-988b5cb005b1"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f84e940d67f6d8dcb4d614c548632e8f3ee49ddf63ccb3a6fb266cbfda0c0d69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/33c6188cc00c0c2304df7a71ee586fa44a66ac13e8d24d1e3eea29e140c3319b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/415ec6b3b2e76522ad643832bd2e3a4f3db35e6c3863001fb14ffc99c4ae0038.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cacf8279-51df-5a97-a592-9d328a81df13"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/10aee42af9396af70a02a024a91e8001bdca664c07feb46d7197839d1154b56f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/536293257375ec58cb8df78e42b022de60f62079b90711e4ac3af234ba14d302.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1dfee51fe009f349d9cadb313a50eb5ff96741d800c77b838bf73df8b131f4f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "aa0f37dd-bbb0-5ef5-9907-001d131f553c"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/29e2c6bbbd35d465cef33b5b971ca6cd79cbfe118a0a9afe2ce2e877000b1361.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/baf7d29e8da754a3c415daf466e027ee9d26357b2ce23fcb7914b0b353215273.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1a6ccd46dcbabe14ce78551ffb705dae68c0d0af0373084f5e28a2fa7883706c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2a358cbf-1566-510c-b15e-274202339d18"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f2aa63b50cf95994e75fad124b0c55c5a418bca1f1808a6120d37e4327fec3b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/224304a82dd4b2f06b4a3f870fcfbe4b810e67cbc6f74f333e5553f297203a3c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/82c3b014489efad40f403d0156adcc855eb4a32b5bc55e1aba4a8947e2d02a9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f300d2d2-eb25-5d81-b06b-2d25fbadd9ec"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c7d9bb8907761607af4cc99302b64f3553a0b701cb4c566222128267142f91fa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f331d708c02ebeb6739a926d4319be4b22c0bc5789441b8729f593f293931b29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a4cc5cdeaa1d3d00605aace4c7dd24b067f4900f3167784269d73947fce3c7ed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8a1d591d-ebb7-57f3-8fd6-23897edf00a5"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7537cdb046e4f68a05caf76242aefc0b4e7cc58b27a31e0f48c2318b7b9c5a71.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/95d72da22e199ef1b0c7b02f027809964320410cbe4d4d5b8640707f08751b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/05f645778c634ce19db256c54839e9b56089b85bb2aa47c2958bafb6a08b22a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8867a364-4a86-5118-aeeb-e8adc8419445"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7060a0d1a0783c0718921c97e851719745566a693ce8fcd4f87e047b1ed748a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a1e801a24950007badf640a64cfa8c01ea20be597ef10e7793af1c0a63f5ab33.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e1953711645f68c272f55d441afc5aaf1ec28908ec2c1e804e8025d609a25ccb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8471e8e5-d50c-51c2-b946-b92561322441"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/34fc84702a6d460be9567df5468186ca5659629843b7787235629cd988f6d623.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0869dd16a6aabbb8b17e0a89a0f98521b1619d47d4a782b6691de770257828e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b16a6d7490b907c424902acb35341a291bc1cdd68ad8a96523ab899fdb5a375c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cd816f58-b2a2-5a9f-bd83-410109d2df31"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c8a83da2c2245b9855150abd033bbd3b3b62197b48a70938ef1db61031068427.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9529272795546acbce9687c9a350d3e6b5b071f4dcc101a368cc55dbb2492ad3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/06841a496ca7a062446f26e8046dc7e4b46e0286fa1618424fb241d0b8b4956a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e2529ef9-37e8-5fe2-98ef-82136023e5a9"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/35158ce3c138ce5caf8986044e6345a4a710a2fe1dc3e4b2947b2cc94a8e691b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0cbbd425759a649c1b083f14818c4694db4b02f3b9a0899d383de8b1d50332f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/591574bf9ac380b1b3777f53a8c58ee456670a2bfce9b81410883a605d69c8f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6c5c668c-0b54-5593-b57a-cf2a4c681609"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6f757e30e2c32da81de5c0af134140c7915ca351932fcb2abe1499aa7adbeda2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1b165b319af7b741710237aa70547c40f626d29fb053e6884d81f92691ffd7d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/586020ebc233f4f560d15c5cbae0870f8f10983dfea11df6a6a6ab60d1c77b1a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "69164a63-b5fb-5259-9dde-e99bac5d08c3"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5ff121ed82e592c1caea50b954471df31907437bdb779bdd4ae03144cc2a3030.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e5236e587ed4e3f48a2eb3f4097d910dcd3dc1a0324e8fa50264057a444938b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/330b7cf6b058bb906d198c8dcddd6ae616533edb1d37f04b836712285a296e9f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "67d71bcc-ecf0-57a2-8af2-2de9f591c6e0"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e4497ad29fbdeb82bccbccd968bf40c3512a76dee23b2996f31e99fe49010401.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2b05f8d08c73b81d59f895c9a910fcddbe203433b4fbdd0ee12e452f3d72b3bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7baea81f11ca52b09be20308e7ee8bdb75ff6931c52a54b9b495641a579d9249.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "53d14356-3ab0-531c-9e45-2d0d3f400f2e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/40f5356b90bef37c3f511207454219f150cef8800394fd113be7926c57ec2ddd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9e41564d04cbae67f045a2083cb323b51b7be499334eadfb2918c4ecc1ecce3d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f1cf72b8a9a5d3bd743f136975d329eda116ba7531ffb51ac70bf7840fa98719.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e4210683-d2f7-5342-b764-9ce466e995a7"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/de2021f3790422767553f51f416030e1943125a9ba512da2a426f881d06aec51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f52f705c45dc045aba2773137d013a71e8495933155ac3bcd856ac92a0830db9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5708506d6b633853f9913a419db29c3283b0c177c25e4962d32fb16078d929f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "93257e9c-2c1a-5b9f-8701-3efd87e45bdb"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cda20d59a973aba2fe589a0e8de65a46f9d6ba5e8f1fe25ae23004fdf9e6d044.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b93c8abc4465d29956fe201bc0c2e96d8d2760bc220cfd11fde189ee4027dada.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/763abcaf25c81728d7d479a1c933d5a985a90cca06db96e8dc061f25465a6981.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b89bf097-d8ff-546b-bdf0-c61be5f0a31c"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f31b18d3b14cc2f21c737f78ccddab23ee2f6b5d75de415b2e3e5255be67818b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/92b8b3e568be20f54eab972148c9a0c2b1cf7e2b0d88fef968ef68d04d26ef9a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/85a46b323f4f2bc401fcdd68f261d05d902896f77accebedbf2875a250cb887d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8e10d048-f490-55b5-8b20-6b27d5367a30"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/88a5963a26ff299b01561278c754c83d1881eac454771554e0fa42c6ea9b3c8a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0b8a93c0a2e197b4ef224ecf43e7d03b6a725ef5564e101386d27aadf27d9687.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f5d07f8953b45334a31a60d3224c953fd81c684969423ce72e1b98aaad16b405.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bc2db615-2925-5e46-8a5d-9d6c3a192d50"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fcc85386e945d391074b9d5058e60393b98da3de42b86f0074da6370b4dc6aea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f59caa46f2a20be420304eedbeb8266f1f940b4595090f6168c5420762f34349.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/aa14026b5604f1247a18389181f9421128e218d1d5d7d2c7a115148c1f973b49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ffc1ed3e-8a2f-5ad4-8413-4ac86f2422f1"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9f79533bdea60c177437f23ebfde99a8a2be9cbfad28d246fe5b2ae3b2803685.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3c44270d4f0284d3a5815e14360691535827ac455e04beb8d5656149be3f68fe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/83aba861b2fbafcfd295b1e89d417c1d7ebdf2fc34345bd08df6461bec482b8f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "492f0684-6444-57cf-b9ab-f56d851b6f13"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 1, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d69b2a6e8d1c4a7be1a784e4a1cccbef87aacd672e215b4e71a571ba9d3feaf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/543de791b8488b4459e1db8262b071c278e1a0500a4b6366a70d4b243d81d71f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5f356397c97040dbcd278e87f1099a7664d1c352ea42328162161477fc59d7fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "eae3dd17-7680-5eda-8915-1a36ed2f03b0"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8e822ec0-b226-5b54-a3fa-ffce82467b80"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/18eb3014119d6f8ec8f0a63634316f7709301d8620b54f2bda80e9712307facd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b3a9082ae9e13443931c6c04150dc1eda30799d8a9517101fbb753a59a2eeb31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "72cf79c8-d237-5ede-b05b-111959bbd1b9"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bf75a207f768473b62e89c333bdd19d18230e764327b293298831e2862f8091e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/06475d91401c2b158e55e30d4e28c2d2107b30e1c02db14289bb34fb1de4cda7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f626ddfe6e12990a7f9aa2be241b2782511313f0afddc8d9d292ca378fee1695.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b00ce07f-9e07-5f83-a74b-54a97fe37b77"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ffa7e6f739470cefebb3de7f4ec57e63eb724a7ff06fa16eb3d10b064c7e0b6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/51506dfbd08cbe8273bab9f8bea73cba1cc90486cabc54f530a4731ff21d2e31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1ee39bed8d2c4e9fdcded560a80a70752c20afd79ab6064c6a9cdd1dffafc558.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bf4a3d53-aa18-5df0-b5c8-8a058ab5ec36"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/edde44749718c84f07d3455839c3d6be5ff5f5a9fc70138f696b991a97114eab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/51977a9d82a52b16956339de1efaa6b5810642e49537202cf3d0259c3335a9f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1bdc734cfd7b62ce75e0886b2127f13c011a4285058f9ab8abde959700afcf55.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "569ab1b0-9ec5-5303-ba3e-c3d7e6a4f769"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/44b6c786dc73fd79dec2ef4764c262abc3b5b177188d694507708003d9141765.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5ae9ffeb353343c24fbccead58d761ba4266716224bb0b41c289395c644033a4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ea93b264d111265059a37eed850d82b76f38932e81e2170262c6b9be3f077253.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "caf929e4-a070-5ef3-aed8-5a916892b20f"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bb50bcdef7a789ad39f647febcf544e1fb2c3c20586fa9c66347aa00b2676668.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/07dddc44df3d1d9c11ea850d1bb91baf0b9d03a38734a65516c9ac71664c4a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d76935641b9cc5bde511a0b0d8fd8ce8e8ffc4508e129f5e55684ba48d476bb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c7c4277a-2524-59b9-98c8-3422125657eb"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ad9c676135a7f311c10771ee8b851f71d0a386851fe0756b7406f9cdf7cd22f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7027e11e3059f4edbafe412fbbcc0853de7df3e98fa050e377a6bd1024fad5e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6ac715409360dd248935328129659f9ee66daa2d79ba52fabd44f16d8c509275.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "df7ab88b-e699-519b-9c1c-ab1b37699c14"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d7ad9fa5e3e180370b7637a6e117dcf2fd4e287a6f11f98d98b96273174b9d00.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/64b0efd6b80c200bab3ce6a00001c2027266820012dff5fe92742d5c374eb95a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f82d74eaa135881138dedf0e3a76c25c6c0278222e27435c1032098da6193a77.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fe5368a5-d8b1-53b5-a6c7-24893c1a525a"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/41a9bed5de1a9cbc2fbdd01435c6b6f3b563c2aeeffe20b2801d6a0ad21accac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d0c6472606d40bad966513f628dac2fe6180a98e068addf98d952da597adeb5d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0a5a3c93562e935c7dc47800070994ac8015282d4ce596fafff166f2239c923e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c8a4b5bf-8202-51c8-9496-a3376915f766"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7480b3dcda0f33530e65c4eeedfc406e079fe6cf93379e926929482105fc5d28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/618eae0d5d24c5c93b57984e4e09d9328cd8cf4d53efd82c5f44cc8f1b366f74.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c31f0c7ec045f39a166135a4041238daf1c97045afbb21f0d527054b854fadd4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9d40f2b5-a110-5aae-a02b-bf27acc4f3f0"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9f581b443e3bcd1709f47421d75b6dc569c92736612c0164658bf6e58a389056.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fd3bc6507483560729a16be73c41683d05179a90afa424c7ac06420f7216915a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e7a13e9cca04997f48bd9955d9dacb43baf2b336bf1c494a6242ed3c97ff5c30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bd58f2b7-9e92-58a5-851f-900227514d7c"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ca121b2661b48f1a9d93fd7e95596852f366e283128eca7a75f341c294d359b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d4cb3cb37d6b33c07ede08b8b4266c00424e1e1adf533381ec1ae9e5de4193ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cdca3d059c6397ccc2de12834efcb7fa6a5e0a74d67b283504d08ab3025d66e7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "cae9e5c2-dc82-5497-90bc-0fb61d58d62c"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2848d21401746aa2ad1373c63f33ec3c2ea5bfad062d11aab013f79e4a3fea78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2049aaff80c04a294513459421bd504089a74841bd3b53654097d907e41b449a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7fd1f64ed37b36f041ddd4cab5e951733c7bd27864476dbfc67987d02b51bed9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9c5eb397-8a6c-53a6-87b5-e36789679d66"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6d4e33e5265de77d1ef561f01b3d559f5b4c1e0acebfbae0d2ec6bafa8800e0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d0939c9c669392d30f4cf1a2ed9a8a62864d271d3540addcae9a1a7965b4a96b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/084d89112edb699b0f794275e8c8475358887ee5fcb4ba8d2455150176d9b254.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bf743591-8506-5869-9e16-be6e8cb3aa4e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9182ebcb9fad60b1714c446e45d03c0ded0a3d2ed03a996d82dd914ca3ff6512.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/428d82040465e1c2886da1b9c8fe8644d46337337ad90962b6fad9bfba1e0f84.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/604b108504e38e552b85c8a72631b88c29f430c71bd346b31f39cd9907b578f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9b9f5473-8f78-599e-af55-b83dff6649b0"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c2ab55572c904578e40853da93c828400e983bd15ff39fecfe25b3026be3371d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1742cc1bcc175810653bd992ad0f2dd3cc17e86d403021d2e2eb2f44d9d56a49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9720e2adbf1b0f48444e0bdd8a5f116b07c92d9a88407428612de9f2baa1078a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bc426141-767c-545e-94e2-4ab7ef1ad6ea"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d24cd69d76773dd3fabbfef2465b5f096a8162615a0988e5c8becd74c67d6bde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1da6eac9d1b4b420ea08ff1332261aab3a4d1a61caffffb098fd60aef8a0181e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/41197000206088fc16da054ee9e09008ec6190c6764dbc20fadf3e386433e073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "f399b280-d1ec-599a-9a39-69f063be5ee8"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e30da1fbb3ea187d76023a7c6ae9388769ad170bd75caff6831260b5b243cd32.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/141700f08548dc2688d216feb4440bc59acfa53ac13909a69e76ce3817714e04.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f8929540d053d9f31eb4c69a9f3f1d633ded71f515e52cdc72d10164d33494cf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "475965cf-cbad-5baa-b685-479860697719"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/77867f4e7be6a6001e1a6fe153212bcab0d244f3f8563064d3d76928e2fda6a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c7bc074665b621a3c61968773af616da90ed5b91b2677101479ecccda1f155d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0a3db14e680c20321dd8205f39bdbc7a1a2faa8697892872dd0fca19828a7393.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f06424d-4917-55f1-a49e-92062ee6393d"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6bb13eaff6f3d096c87a78d922311afb9ce2345123d8e625c1313f80c7c55701.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c649dba72663751615ec7e678ae82d6f5bd968721def2c37a8563b74023dea1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9c4d35f7aaaca8537776cdf518902ff529df766b09269a1e33157d5160e4bb46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "56cd6408-c046-526d-81a1-7e25013dfe00"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/400f4ac9d6b37f5a2dfd208a7bf9d3d76cf684cb55f5304314d119382d221867.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/88a2da74a9473f51ab4bd0940c2419a6195b523ab07ee39275b2638aee02896a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/03827c56cb7a7e6a56824082ae273090eb683c808fc6fd1e4ca1f456fa55d670.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "3e7b0302-b42e-5674-8b89-9c9bc8cef185"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e4ccdf37ef0ae6920dceab50ef6843d48bc8274de49a5c8aa1928563b7c6cea7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0f11d8e37edc1051f193d302425644b879c0da5a717725f7f6c778b4db6e156e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5528a77d13ce952af933c4104b0e8500034750b97502171256495ba38770d77a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0c4ac2c3-4269-569a-a7d8-24a40c88d6c8"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ffea474e7764ebbff185dd61253ba800b1f42f826af3ae697e3020f36e25991c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e10a6be5ef3d2fe137a102b25f654b51575af0827ec97005551948ff1b675d13.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/12c1fbe083fac133d3db40447b8784302b582b09e9da48c10d8b2296b705771e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e251f5a0-c781-58a5-951e-21de1acf7668"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0257a2b500f10e52f34de91a8bf7ef1c2c5c1a114b2965d362e98a9ae733c594.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a5fe9684453f3782838f306ee5735a2d0df8b60ff06e3b152babfc2c9d0d6cff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/49f45fcb83a09b7ccc4d08e58ee430b9c14a72ffbc1f3532be08faaaf687d7f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "1f9ae2ca-2f03-59fb-aa8a-bdaa967c3513"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f2b17ac0105f7f4f222db1ddfd6cde19d21e13787ff77ec8fa949faa5a2c842a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/dae6694936887a06baf5aa04de4c7cdb7b9d82c6bbaf9ff3b4b378a9cdca4aee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/95db2b2604fea51168744effdaf40ae6ce77f6ffc5b5de9eafaa976b4b4711d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "12c50ef4-b8c1-5f90-88e3-8721a5aa5c70"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fcf9f85e825af47deab6a368c63557bc3948436de797b80c05ce6da67e5c2e70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c9f98854cea847424b47fc7f3060f0fa8387bd7aa22b3c354662e41af600e38d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/da97689a1df734a149bb7835b7ace1eb5dc0cfd9668d8a4c55b9a0707b7462a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ac82acc5-5fed-5aad-aef3-f63bf7050267"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/85f1751a221f0bd4d54a9f035738c2198d50595fcf3abf3e3f79ec6f7f8762d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6eebe4afcf5544688fc3ccdf988d5339368808eda994c0cdfb0cbf04400ed360.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d33fa6c68a50dd88a36b42be9b708389f15275d3208a458a5b3d00022fe18101.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "35d00eff-429f-5302-9ebb-a3d1971a22cc"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fc8991c18d4146514c2a2eb00d9bd58e51b5b390e6e1114d14650d4fdc87e0b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8ed82cbca3be1b2279cbe9877b60b0c78d2729abb53ed5d42c004918bda3a24a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6c1fe649f50ea1bdaf2bbe3152fb6cfdd63350ddddc95bc4cb679274d1c1b52a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b33ed0a6-33a1-5651-ae01-36b17523f6aa"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/12c3a94a24cad0a31d238be224ac57e308438d05b69e6da31e95b6ce48ad90ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8f103729fd7b094738b57b5fff7dc0437ce671319a3fec2c99ca2288b473b6b0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0f369afcc9094d24a432f6d083a44b8f1552f1df9295fc354450d33f6814ec30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "acabbb4d-4ad9-59bd-ad85-aa59e5c61616"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e52a142fc3ddc699ba1237d1650dd3c572dee8e41a3be278003a119f63806272.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/db0ee61e3756d7ed069c41c0c4b7f25b128be64c02c67226de9758ce41f0ba8f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ff0b80f58c536544fca0c4698e9886791835a87ff9a86393a5796d389cc3181a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "eaf051f7-e6f3-5dfe-8883-293fdc53d10e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f2ee1dd42be5a113982deeb1e25ffe385317bcbfd5f0861e5c65da4936367269.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/daa86f2d70f0e717a94cc0ad1c48ec6d04ba6f0bacf9147b6b13677156f4e4b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bf05b52642e922cf6c8aaf89e9912c0ffb039bd14f6d263c1ac55ad69251b0c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0aee85e5-f45c-5574-ba1f-84fa3e8bfb6d"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/64cc8eb2e4ddfb06f8926d5fadd79b08f8768cc67dddc3a9bdba0c72e38c2eb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8af5db265e77eef388d10de68e19e9283f4a3751335874a0bf724ed7f9b014e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/83c5a23e6f800854993826d8405aecccf6ded4371eeadcc812ee2a9432cb8bf1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2c6d44e7-ea42-5814-8813-6a8164649bc4"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/d12ab6156805aae4a032ea734d04d40887b55cba30eca0bc56292cfc8c35c94b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8a12b6970ae9931ebb95598ca3c34cc39734dd9399d65f24175684f9e930f293.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7b13434e242ef299b10bc01bd71c891b0817563dd1adef4c5e830652a7d67793.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4795bcdd-6f37-5076-9404-a44be707081c"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cce3a908f1bcdf646401f13cbba6af7e4740fff5d7f7bbbb4d141ab6386fe5ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1b74de2588bb494ad42971e6539b0d25b342cfa34dea8de7bec7f85cc1d7dc0b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c6421773900a8f532db28b70be9efa8e11f702225c1dd20efbb068cf302e9e9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b88e2f5b-0006-5a99-a92d-cf4a8625559e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 2, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/294230926a74d901afd425124938441f7ece0177ed3f4974efa259ae10f9732d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/17320674444b745f65135a0be938b92cc0dc07f3e1efc43307070c6e00704e35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bda88ddade862721decbe33e68b2fbc2c9455339a253ba1496e10c51574aa54b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e0a836db-4eb5-5404-b90b-feae5f3e075e"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b6cfc78f-2787-5033-87b2-88ddfbbd32b1"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ae988352988491999f6b3bf530fefa54ccfcca61388316db2ac7143d85b81617.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e65cfc5a4ac5027a5919b917fedfe32ea0f37f502e4fdcf512233b54d8b14831.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fa5a7eb1-1084-5daa-ab0b-72ac0af9e44f"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/03a59fb5304816524384db05153013cf9492d66c8f38c5c8e9790d60ae0e22e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/41612c3c743f434bc2b7f03a410a081ebdab7252fbec0a3632e1dcab5fb1e0f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/a1f505419574e352909272d38115a87b8bf11c64c09c0d2c895ea1357c36f830.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2dcb7d09-5ca5-5b35-aebc-33459f3ca110"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b34e56ee8ff6e6506296d3cb224818a7290cdd862c26f0e6bda8dadaa6606932.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cd939d326d951b814c4934c4e605da9937411eb310813d374ad7f0c1c04d257e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/29f9740b01887a0809ac685ab528568060d8b7c708c95a401f31156453121e75.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "8828a62a-b05b-5c5f-b48d-46a0dfb68c36"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/dac1e023874b9ca3ee2c42e26bb20310070f35d6e17abac89b146a671df94cd3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0e94acf0571283daa1e5291ade0e7fb2169db84af1eb6a551e7ebea293b8ece2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e36a9772680264c08c2ec945b5cf0161bd3fcf878eaa49300019b14dc455d241.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "abb09f82-ee47-528a-905a-f86b658621f5"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6df3d1dfe139dbde80d0f2fd39cffdef6214a541b98166a440c1e634fcde308c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3e1a21e07ec269b3c844ed55d3a2ee6041bbec8c9bf398ff91aa1a489d060503.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5b3552c01873ee404f793c2a4f963622c0e32354b3abd6598143e5744e7697cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b204c6bd-4c17-553a-b6cd-9b2be4bbae28"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/211c4895dbbd72aecd30b2602cb170ea36103da87719d2a45c4d1bee02122069.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/44f48e705a9d780ea58fdfeac1becd8ea424a1150e6bf2cc3a296c70b6ac7bf6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/49dbde78e5eb72cd10d165a7c9e81b4397785c8af20f966e22d03244daf26d09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "05a21a06-56d8-54e4-b6f8-18ee1a3235fd"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ff7c352d057b2caad2211f9f6557cece97ff78b0cfec6f93cb1a5acedaea2cbe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7e6f2734f9a9dc2ba823520e2aea3fda225e7d7c67baf6713ee9a4ec7117b7b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f7bc26bb13c7ce97d30b957b27428ee339c8dcba58a01726d09ad6afee0b73df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e18b75bf-014a-57d6-a428-42f15d177aa3"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ca05093d5ff552c18c0884eb7cb0def6034f828afb6874e45180905464451d08.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/be0d9bb1fcc3251e5af541a3ccc9c408cbcea30d2ca897d33f505c70166898ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/efccd97309812c22ecdd3ea177de3eb7d320fa63bcae94df0b36c01f97283143.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "86da2b5c-41a3-5409-8bde-0c0b85162668"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1df82ee94789fbfe861ca1390914fb82efe20ff574848bdfcd75db1a8807ecd6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ffb4606106fd02c7812bd730d0e0e060861ac0bf83979684623047397608afea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8f296025a2913b2a4b8229be5dd37e2e0f891e9b8d9491158f8e9a772f1b2c9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e28ce398-4258-5170-a14f-a247318ff3d3"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 2352}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0287891654683ea3a2eeb24e855d82c592586a46ef345b85b695ea7da66c1ceb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0b75cbfb0b6384af1eeb044ff5169e94424ca6f4ca8a1df3839d7e51c7c71663.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f041b389a02e17e8856ee38acda4ad22f3428813f8d1f0a41a6e83e3bbf08195.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "efa942e5-4b8c-5cb5-9554-232fa3b7b889"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2352, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4e2f7c09777f8f3a84d6e029be95f0cb1584c2189248c2a4387aa6d7f47d6610.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3c7b7c7c52fea6a4bf151c146b86b4082d4ea79cff7cb62cdb38d37db3c206bd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8148435d35256841384d8b053eb401d2838324ac4f8a9dbbe9b00c480b333d79.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d3cdef82-ba3d-5778-9b1b-81864342cfa3"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/6b779483a6c9dd5487438e8e3bfd0a3f44e6a413f9f33a70083a91769d5605bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2e7fb8e26f2fef885dd59bcdfd1e084e177516dc870485516077f1dace56644c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/452e319db2cd96d2f0bd19fcacaf3fd5740d6cf9e7bb64f3a0f7895aeba8955b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0dedb1b6-22f5-5317-8887-ca42e277b7c2"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/68b92b8f9870a17c46e57861758e46ac1b433b18b64b149eff2342f33fda74f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8aa4c82c0654a6b958e1bab96d235f85fc53dc504a1b31beda457102a573459a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/875808b0a140ddd19e7e361bfe7c617c6d7119d411ecd32bda7dc87db9adaa22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a4fe2d19-6cec-564d-a702-0312dfa23b28"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7280074c2f757a00590f7ca53f6fe07181e4429a2e27148ed24362e5a657630c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/66378212c72e49125f191f66575820ecce1686e4c315c26de78ca7b47219bd4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0d9808647525393d405e8828bb0ed5be4ed0fe3f353da2a3021dec6c3e492429.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "d74d520c-1fff-5457-93d0-0301994b6f31"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c798a0e7cd9b6ce47079af744473546736f8fdafeb72dcde159fa46e347773e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/21df3f2802206b422825b95ed66e82c77042c667e5583c7bd6d1633cff577a67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b3d0e05cc6db9b8d8d97dbe71d57d072fc3409c97f8f6fef755d77dba1aba4fd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2f44bd82-037b-5ad2-bc34-88f523346678"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/87d6c96e09297c826af179868b0845aa477aa589457c0836c074874b073158f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/cf15e984577db3744b2530b33af10b6991612933738edb81e1eb18bd83428b4d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/00e4c13ec2a43ebcd3ca1b5aa821602aa718d3fb9b07cc3b3f0fb5e1b4014e2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c2e44540-2f70-5245-a34c-2b91bf89a7ae"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f4f283e94570f8550215a5a8204c7aa91c463cd7358449a67d435693a6be2d28.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/526b33cf474d60085f6ee35209b2ddd9d86a2e603c185bae2c88e0eb5444c7e6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9287f11be8fe89954b7268b51f20c39a0c9cee94cda3171cf8f12cc9470e6f97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "9109ad43-fa43-5892-a18b-3351bd60ede1"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3918952315dc8683b0a0314bc3e101b46d13eeb8e95f07191052325699611c89.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f82776d38696acdd3de32a18296c6d91efdd6269c65f89f9e20055e3ba091bb7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fc181f048ba6d796bdafc0fc668246b765d58b1ab51a5877e8bea90ecc65f128.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c1cd8d74-fd40-5256-a30b-5e1cb4ef5c52"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f3d39bbc27ddf3e08445c283db3de4af382e6bc6cc61965f7d4f617304c1d429.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/dcc58dd550147875389908795e8da8e096fa86f23fd2558088fba80190c1adae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/feaab609698a90e78fbf01041860c5cdcb020b618491ed455d1a5faa94ca9a33.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "ee702d9a-f86d-5b59-a5b0-e6b421997b87"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/5595c8485c97f786d65e9d8a3406fc385c18b71014cdeeb71a2fc884358357ad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7f456e8957e7081f79e3a424bcb40420affbdc3166b7dab7b8ba111fb0520013.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7a647a19139f5fb7c0835d626d7e46528d525064503a8e940903e49569bc70e0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "1e22f1b7-23e2-5dee-92d5-f74e6c6ca3ea"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1cc48dc5c8ada97b2e2ec04901e9bf3926334cb57a04746e6f2df246874fff5e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7e5683fd82467a6f1802acd250977c1f0aad29cca1fc618a4f2e3c6c01f5c666.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e5591ee776e2957a3da060bf983f0af08068c57f2f9b58b8efdf8399f78583df.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "e9ad38c8-2a06-58d9-b32a-a96e3255e630"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 51, "Nq": 1120}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/b08564679681a8bf55883e76d236c4292dfa1d37393decf6fc46aa8ca6767da9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/89b4cd64c6ee392641760d8bdb6a2a38de6e3169a5d832b8dde79d9e16728f40.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/08541e29192c85df068cf916e713089dbcbe237ed56e40afa6b7da691ad6d601.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "645dc728-94d6-5801-a837-9a1f4201b237"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1120, "Nq": 51}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2bd6e5ef1203da39a0ecc01c189a2fcd31bcdae5dce5fffaf4853801bb7c5ee1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7cc0a5a62b319d549add769bd476f7585c2b4f8fca374642d43844dd7b43bf63.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/627d280ed3a32c51785b5579eccf43f09ab13095d4b8cacf1fb8fde7c27a4a10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6fccac8a-d4ec-54e4-99d1-48ded5d20120"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fec13c7603a43874169d79205fbc66ad6be691a781e9ff39616a1a5008e19adb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/9d911f2b7e2eddd683db2417584f60c19bded49950f50b905d00c8b232ea8c0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/7a9c3ab7f9d138bfeaf395732f8d94a225f47b464d8c66ee7103189d4779ef8b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "29041579-efc8-573b-8642-1ea8af6a2cdf"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/446287320ccff83b717105a20e909e20bd49815815bf39ef8e5476d0c37dccb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c3b9e63bf2e7e5ea60d46d5c967d5ad35918d9cace54ae37014198f19e09ceb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/420d85877a2f5de4b505fc6c19303be6c9670639ed3703248e96786f0448b65f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "fbb0bb64-f32b-589b-bc7f-48b571a90daf"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/70ab0fda0bca48e20227deef9efad4565db1a1e36c1108c9da389e811c13e44b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/80029a43c39a8490cc03c9ec41e44c2b5f7a72de751b180f9cb337d2bb137a4f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/db2886952b538368aa184d670824d7d8993d86f5ef765eee3e9b5c38e79d602f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "4e69b9b6-fb8d-51c1-b88d-d0b346d8bf92"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/30abc412af742c61f904c61e08c2bf76e46a80b3ec53231b7c3f52d33b907e0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/76d72760354b327643d2fb91c77641db8dcd25250ac99ad0e165712b9b33c02b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/27546ff7f9601fb6369f33df086c24dea92274c1930ece92104e2101dae54ea4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "2b784abf-d74d-59d2-840f-d750c00907cd"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/8e196cfdb8d1954c4989df6947d3ae0970d7d6216d5f82e62045ce36e1166116.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/1fcd744e8bbe80d4f7c928f179cef7d717c8280d883337ed963db06122b4bebb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f53349ca78e4069a1f438f640947a78cc1758de55d2f8a54715f7ad2b95eaff3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "b19d6db5-9600-536d-8f4c-543dba8b95d0"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/4d4ad61e72b663e99609205b5ed32f22ac4fdaf42fbe6730efb43f1adb928583.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/3c30f7a4b43e36dfceb650309d0d8fdfa8985ff5f960f18095e9c70e8f836015.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/916ff6ff2676cae1123509095fe2eab9b02d67054910d89712ef482b1b752579.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a5cb657f-b218-5f59-b04e-23be91dee147"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c2682be6c09bb4027a65e1b4b4a1f71f34996ae1816b54e1a8ee844534e2dc7d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bb9faab5fb8f83f0c69868f36ab06cdc1f8f8d12fb050a91ce629e2cf38f1245.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/fb7f5265016e32ca1b1d66dd0b0e27d653a612c93cea31e0b385cafcbad619e8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "6ac66abe-111f-58be-8c6f-51b54e0e7336"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/dabdeba7515ed93d7d3a31fb001daa087fa46ad9bcddb905251c4ad85f5f31ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/ac0e420a5f115f334284d09f1eeb47c06629d4753f7a36af09e2572b21acdd38.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/325a2cebd9548be6d96fca378e643666cbbedb2bbc6e3e5ac05de557f16d1546.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "365db76e-b098-5d94-84a6-2e0791f4ea86"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/bddc2cf3eff363281679c7aef2c854f7b40b63647741c72569d7335c60ec8cde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2e3db22329e5ab697c1fae0fc8e00380f488dce77feb96ac7d4c994aff833fdc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/53e0a52e1535d75a83beee3d26f766cb9e2d680c7552846d70a386876035dce8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "a656cf2a-7244-59dd-8937-83cc198c0202"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 1024, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/eaf8c6f6d6f76de96ab37d184fa9da2fc3b7ed65afef52e74e5a135025655102.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/30fe75c65cd95cc2ada0473892ddd0ba195fe6a10155e40d99fe7b783a26ae88.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/f9bfca3e17c254769724063ad5676195ec838a2d4a39c617462e8cd6c3685263.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "bd1288ed-8967-57c0-8c57-a6a5f6194392"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 126, "Nq": 2560}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/0f6a9151c12245f1020bd27e1560b2223221e0c737018aea1b48cb0ef4014f47.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/c0d3863f20d63cee5fc98a10aa7c7b80ce646c521eed58607a2d4661adfabf03.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/aa55e8951e74ba972b0772ef2f0d715d5d87f46184d1727d658b2000afffa58d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "0f7dec0c-09a6-5895-ad37-d817ddc5ea66"}}
+{"definition": "noncausal_dense_attention_h32_d64", "workload": {"axes": {"B": 4, "Nk": 2560, "Nq": 126}, "inputs": {"K": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/819449fb07cddeca00463d5ddd4d39eddbbc6c0add5fc6865b219aeb32a163d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/e346cb9b5cd501142ddb0957857822280d6f55dd7c177a126da9da14e3011615.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/noncausal_dense_attention_h32_d64/2ea3d0ef00c7831032dc328d6d94790201d41dd868714f3ce20a972b4f96b3eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.125}}, "uuid": "c6fde8bd-ad59-57be-bd10-75481240b2a9"}}
diff --git a/workloads/gemm/dense_gemm_bf16_nt.jsonl b/workloads/gemm/dense_gemm_bf16_nt.jsonl
deleted file mode 100644
index 3985f9f64914c4f9b2b73ed58365aca65ec6ba44..0000000000000000000000000000000000000000
--- a/workloads/gemm/dense_gemm_bf16_nt.jsonl
+++ /dev/null
@@ -1,93 +0,0 @@
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bc334c45-f866-5289-891d-4e12672c7e97"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2a724567-1566-5652-ad8c-2546de654e3c"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 12288, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 12288, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 24576, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 24576, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 6144, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "717bab44-772a-580e-bd31-8f449703fca5"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4096, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 12288, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 12288, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 8192, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0882664a-03ac-5037-b167-fa264ffa0639"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6c24a9cf-df14-538e-9876-3431874e3701"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 24576, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 24576, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 4125, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 4096, "M": 16384, "N": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3f471772-09e1-5090-a367-35799e617731"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1ef85b69-b938-5554-b019-1cd333110897"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 6144, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4096, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 12288, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 12288, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 8192, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 24576, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "89b32f3e-3660-5591-8885-48dffb786922"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 24576, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3589b4f7-d841-5091-87f0-84fc90925259"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 4125, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8344cb3d-646a-5983-955d-aa29071bc411"}}
-{"definition": "dense_gemm_bf16_nt", "workload": {"axes": {"K": 12288, "M": 16384, "N": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911"}}
diff --git a/workloads/gemm/dense_gemm_bf16_nt_n12288_k4096.jsonl b/workloads/gemm/dense_gemm_bf16_nt_n12288_k4096.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d407fb50f98a3e3607a33387aaafa6cec9ef8cec
--- /dev/null
+++ b/workloads/gemm/dense_gemm_bf16_nt_n12288_k4096.jsonl
@@ -0,0 +1,31 @@
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f627f3ce-06ab-52e0-b919-0882bb98ba87"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/dd2c3c9e4563d68e0eceaf6b636447bd810f8ac4d34ea6332190501d8cdaa355.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ff62c8f2-f87c-586b-b295-7d130dd1a037"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/81a96a69b9ff8b69ca337c8723fc291b117001bd7a52afa4e90ae0fc959f6345.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8cd405d5-d44c-5568-8ccd-4aae002946f6"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e033a43092ceec91d029e402f4d191d8a3432a426647b24b1aac8134bc95ff0e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8cd434d9-bb43-53b9-b795-ad7efb0592bc"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c97bfb44e2775e0cdaac962bfdd2f877d4c99dbc2c9678f2128fe8ce562d2021.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8bc7ebff-12d1-5ac5-99bb-d18a9cbdb505"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/cec0f1e9b60d0690b3be938c39a944172aaa7dc4acd52041b0ad79454341b61a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "87ab8f00-5cea-5f74-8f0f-afa1d50d4a9f"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8dec516f-2b15-55a8-ab77-40cae911fa48"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "717bab44-772a-580e-bd31-8f449703fca5"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/3458abef18405c57cdf24fc03a2e5df6431a736fcaf61be3c80faa27f71c8cec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdaddfc3-abb0-51d1-8c0e-560907ff294b"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2a48c351e9371d5f66451dc30de4a50c73c4465d53738d94662926caf166bd4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d307d571-5f4f-5ed0-b244-1e87b6dd034b"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e42dfaea-65b5-5490-867f-7e1dc340d154"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/2ddd13c9a2e33a49f692aac184e5e81b5e4d31735f79e9f98222723461ce6905.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "05211974-8e1d-5b1a-8bad-b214cee70353"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0011fcf356b915718dc22610388ef6bb0a06e952e61cb610954122ea6d9b569d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/0b9f435c086d799e64946d2339650d012947496eea8fddee2c9a74e08a32b80c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6382267e-43b7-5d51-a475-3e153db4dd4c"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/94b04f641e312fc292d4b3a9b930f4eeb6cf1318e34029e64abb117c8bc459a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b571dcbc-df9a-515c-90ac-93ba5e03a0dd"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/c2a2a2369fe08707c7349f0050ff7d8988da2ef92572fd0266bc4ef210e558bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b9572ad2-03a8-5830-8ded-11cb591ccd78"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a437b5a4c0b7c610e16d9828304d4ed04567fdf4469af390be3bd095aa716980.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d89ba855-bc0f-52a5-9825-2f44455164ea"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/ddbc841406dfa1cf43db45375104bc10436d822b7bf5f79acd0f349547c6b5c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1e9d5954-f8dc-5e61-b3e3-4a7adad70ad6"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1aa17838de8ff474520c0415b52de63cf7c284b2524dee34d2f518648193bcff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "15ceda36-6fd7-5ea2-9889-1d88eb4495ab"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e2d459ae-d79f-575f-aa23-3fdc5824cdda"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/11a8b9d4110521467731516b88a4a1f59b0485d642355b10eafb8705f005277a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "cb09899b-2462-5f7b-a3df-1b8bf624f35c"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/67e37cb856bf62111318b054e45b50aa09b3037acb24eb5ef56b9620cf6e3762.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6fc5035f-9974-5364-a58c-0c21a341f149"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e036a483c4de8e0799050c512f59e8c173dc2d31dc95fcd0de9b8f11fa871557.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0882664a-03ac-5037-b167-fa264ffa0639"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/613bf537e5c647929de38c09f005844ca81e16c16bde1a0a76a7f61d16018b1f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6c24a9cf-df14-538e-9876-3431874e3701"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d6be6a755b2ab6726c360d5b1b24de346f6e4cf020be805371ab2c331a730970.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8511c9e-edb1-5002-a586-b37dbfb060bc"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/1848e8723c3271bb58b1899716b88fb309c1947013dff24de9d8697e52dffa45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8c811c7c-db79-50a4-a48d-c1e483e9418b"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e109164e3360fc41f730e76a3c7520d93fbb3dbc57d2fe7c3b0d6a35375b293e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "304cdacf-4248-51bd-b9b7-1f2bdcde0405"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/d5c44dd34082f0533eeabce7e97d842f0f2d509b5430cf3925ebd7f973cf6dad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f662784e-03cd-58da-8401-07b85a5b5e2e"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e874b83d24d3156274e54a884cc411cce04ee0a47587fb5224a3eff9475554c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5319e423-c7fa-5af2-a356-948cb73ad345"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/700d12ad328e6f79c703226ee14f6192ee4b82d1b9691995aa4c10beb31d86d8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "def17f80-03dc-5b52-bf7f-2ba88e69d105"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/a42766eae27a05cd0cce5433d112b0a4de66a8aca2c115f1aa0c128d6963060c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f47fbafc-f0a9-518a-ad27-9ed12d8832c1"}}
+{"definition": "dense_gemm_bf16_nt_n12288_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/e53091e5fb897b175326629f99b9501badbfdf509dcef1a2fb89f27ff982c630.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n12288_k4096/9d053c114f2b2a56879ada915a1e2eb6044885b06fefcda54ef0809e8e551c35.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3f471772-09e1-5090-a367-35799e617731"}}
diff --git a/workloads/gemm/dense_gemm_bf16_nt_n4096_k12288.jsonl b/workloads/gemm/dense_gemm_bf16_nt_n4096_k12288.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..f32ee4777ef21da3185ddf079565ff2c3e8e9472
--- /dev/null
+++ b/workloads/gemm/dense_gemm_bf16_nt_n4096_k12288.jsonl
@@ -0,0 +1,31 @@
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/1834d8cb0221fef59dd0aa037bbfd38a1fd686988efc40f374ac55c928d087b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "220902db-7e9b-5d8d-ac0c-4541ef4e4bbd"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/42301ec5f9595219dcc3beb7e95bb5db76a712f290b969805fdaae2a36c64fd9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ef54aca-2a99-52c6-9c90-3fee81788165"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/dabc10b64650cf13c9ba16f0d658a8f6ed08db14debbb08dfa5a4110027e6fed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1ef85b69-b938-5554-b019-1cd333110897"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b9bcf6d86365b2914d002bab96cffade1353d68cf8374352325d47ec0943ba6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "17f61747-5e4c-5e99-a7c4-f096b147fb60"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8c26c825004f1ccf9585d54f9643aba70a6764ffd444d8feef1502f417763254.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7c1437a2-685d-540a-b50a-04e060aaf3d1"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c5041247b342ed5d1e87984607db331205bbc3626ad61aec9b35afeda20b1e24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "37051054-7c40-5ab2-8960-1bef5fe3fc57"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/8bd318889481193fde47042262404b9013805ec2fa047dd5f5f244a3e6847605.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5a287253-9d0d-5c95-9930-90afbe51825a"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f49b336e-43c9-5d0e-a40d-277dc443e7e1"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f1d436142bf57b63cde785a04b8af90605d6f6d00589f4b864835831569398b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "09f4ca4f-889a-55a0-ad71-c15c37f2a4fa"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/2e7bd8b915ab96f3138f8bc79bc660568ce96bdb30c3e6be1052e93efb3e2bf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4822d7f6-4e76-5a5b-9e53-149df4f45997"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5206d93cd32897ed9a5ad27874582229aa0ccb2be4fbc4c85f109094df40ece0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c8c4d9a7-ef98-53ad-8194-887bdb36a542"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c4195602577fe02a3b08da774f0a1635b789be72b246b0840b0a96b4511e058f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "22680fff-4c0c-54e1-9bba-ccdac73a7f5f"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0f64346ca9d87aa852bd57168fe058aca8ffd2de0296bef3cb6c97b945fb7c07.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ed5569af-287c-5fd3-b99a-3f6fd87e2a33"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/ebf4cf857f2fc072600d5c7261f6603cb8e148ccf67375932d084cf6c8d392f5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a870fe23-b82a-5273-bdfe-7d289bbe9537"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/084195e09b36f674af9867a8bac1eb8582d016d07b020084e4cb1975150d2794.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d26eaeed-136a-5399-999a-baf0a72a8818"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/edb5cd16828e69a470d04fe7d5f5945a5d30221e202009bdad8dca6cd0c4979b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f53807e9-dd75-5cca-9c00-8cc6f48f3a3a"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/0abf8a70022c1d7b2539127fb935f9116888180109feca9711ba816aa37c2ced.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8886990a-4593-57b7-bd60-9c00e411ab13"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/daf3e7198731fe8f5c1278581664e02142e7a793f94412007d6fc74243bf7c2f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "857a2e8b-b80a-5f8c-825d-ff2fe2edbfa0"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d2d098b8-07b6-5ca5-8a79-9f6f7e8f93cd"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/47433137f85c9065dcae604733084b90b56ca3040ed9930dec3b6028c62bbca3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "acec1a44-43f2-5f4b-b13f-237d2e82cf96"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/fcc5e13abd1e01142d65923468524e1968e30560e8c5422e248f5303c1d93a34.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "910fa072-ce50-558a-a4d7-f6b68ab8f7f3"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/b461c1c79f195df83f97be7c661933f70b28d81752f88bc10ee0578d278a000f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d301ee66-c6e8-5668-aacd-b5c4838e6bbe"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/cb4adbd36236071250d65f74ecf99572ebbe8b594b1e6b5f8bb23b96fef72142.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a5375c67-0c90-5d89-9dc9-808e7ec3b6f2"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/189ef602b014081add1f26f76c8fd630a99796c7accde2ac863190b71efe032a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f567a1a4-8780-53bc-bd48-6b7d561d8eb8"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5ff9a0b83c399b2f78eb2b853e251d6b35ba4b31b316689b9e7da724ac62c069.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "89b32f3e-3660-5591-8885-48dffb786922"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/df10bad3245b80319fa29115d7ecde078a3bbd103172a327abb228c0285dd12f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3589b4f7-d841-5091-87f0-84fc90925259"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/c6c0e4ee7bfc3dd2f6fd5261638947a6c9dc19b57c84e2f804f73aee7d6af2fb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8ce49068-3af8-5772-90d3-913923d2cea7"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/f7b1b1fd255f334e6ede9485f1af5efa099183db2a67f05deb007acf26f13f21.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a5a627dc-d290-5391-a62e-6bad43f1751a"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/e4e42b86a764e1eae1fa81d656e8c5e102771f60757c21caa55d73e5d6d77d94.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b630cff7-4855-58a6-b355-47e6dc0d15ea"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5fe060b42b7a0669e32fbe805f3eacf6a6d240a39ff02d7ffdfe3c9f097dc4eb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8344cb3d-646a-5983-955d-aa29071bc411"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k12288", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/910179beb25367eefb07bafd1d103d7bae31450decd4c9b9a9a1cff4ae3ea14a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k12288/5f6caa5eafbc51c553302a23fe6bfe08c15f621beab45ef5b329f5eae23e8aa0.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9ff2a4f-4445-5c06-92ec-f19bd4c39911"}}
diff --git a/workloads/gemm/dense_gemm_bf16_nt_n4096_k4096.jsonl b/workloads/gemm/dense_gemm_bf16_nt_n4096_k4096.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..80db75557f24f963db97c2bfdcf30a77de7c86ce
--- /dev/null
+++ b/workloads/gemm/dense_gemm_bf16_nt_n4096_k4096.jsonl
@@ -0,0 +1,31 @@
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f9d01d18cca7a3a29b4d336bc546444708ca2a69e3edf6ffc8b2b21cf8f9e103.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "491d8c98-25d3-536b-b156-340c7a47c2d0"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b7ff5252b224fb8c43296aa0287879480bebe54b1329a2b2e6143734767bbd66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "59e21fcd-7080-5d70-aa46-c4500984dddc"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/48110da6e419681a8dfe9bd0138c197a423f393932f9bc23e6860c98ae9da8db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bdc8ab37-9ff6-5499-b47f-64d3a6e67f27"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/7fb780f1f85401dc888f50148a7ae54ef5cb9330add32cb9367002896cc43b12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bc334c45-f866-5289-891d-4e12672c7e97"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/746140e164638f316e3f7194af1253732e8b9bdd8916a1ff2471c3175dc470a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8692d25-e816-55f5-acd2-ac13d8e72886"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 6144}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ac55680e4ab4096fba26603a0785978a416ec6d7f65d0e18dcc6dc4a5f804046.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "eb7c6323-04a7-58c5-8478-c695c63634f5"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a5941ee-6c6f-5312-a860-d0b26263e260"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/cc78db91bc5fa9912177b6d04797d20fffa15c79a723cc73bd734e0da54bd068.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8446920-1e2c-5681-872e-2de1f62a1c73"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/9ced1803399698b8952e9f2740ded65bbe10634c2a6fe370d5159def14a77551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b2595a6b-c707-5c07-bcd6-07da2b26fb72"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/d6de65b811f6948058b7367384ce97a3775e91e30220baa568500b0d99f53efa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9698844e-f86b-591b-a7c1-fcafc8b838fe"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/aaa2ef77ebca35203f9e5dea860125b075a7bd9492fbce831f37259f97f792c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "77d0fee5-a7ee-5516-ad16-68c0193e9fa2"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/1030b6b26fb62493c3107544795a18e6d99c3e4c9fb980e28b14a1ef553689dd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b6a67dd8-dc4d-5060-9d3b-7568993299ba"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4096}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/8b2ac9bd31d5328cfd29ba74cecc2bd993882773b2babdae339f7dbf4fcf5080.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/548921500b10f2089b3438da26e527aefb32b759924daf6e2dc361a1e76d6d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2a724567-1566-5652-ad8c-2546de654e3c"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "08b43bd3-126f-54fd-bedc-24656a3f3ba8"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b55364b2a1805bb51eeddf5c00d6e4ef9c97ae6070946437df2aaf5dbce8e3fe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "90021c8c-9d0c-55ea-857e-e12ad4fe7979"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/622168dca57ac4c0137607978c797cced0218d0d071dbd98ab1c6d4611524eef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "67b173dc-ff53-562e-be4c-8e5d7b603f06"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 12288}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/c204f75649e9e63e68d259bf6367baa1ba1b01d6bfff304ae12fe9e47e201943.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0be5412a-a24b-59e7-8cc6-ec6de9b0e6a1"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b36852c0-2e4d-5b24-92ce-2c46d5fc8c97"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "30a38acb-ba44-550d-b1ff-48921f60e605"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/88b1a544baef5c471d5d6fcaad810d9d334de5405f1ddfecf68a65200c5e374e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b1369ce-f541-5687-bff0-28b54246e54d"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3d2301f7e426911b7675e834fff4d4db45715878b803a171661ae1a56eea8609.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b4625364-7ec2-5ca7-9997-48c6aac77557"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 8192}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/f51681084afe525d2afe3aae517ea7fb8a5c238b9f6a6f5b513dad9a60ca639e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "eecc9d19-a7b9-59bb-bcf3-12ee8da5d8f4"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a8fd528-d8a7-533c-be10-6bbb74925548"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/e854106901c137dbef9f2ea63d310720ac5f7d23fbcfa150a365c0c139c8f4da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4995e0dc-1c17-5f42-9530-b660dce2f09d"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/b137e2824a274f6451b088327a16cd352596a53126e16fdb48a087bd36c41f00.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "54ea04a7-b4b5-5279-bed3-34f4ae1cf45f"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 24576}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/3577adc5f6c646dc329d6b542376fa58eec21da21e5eff96d3d8fe6b54de0e80.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4d48e098-93c0-5f3c-867b-cb99441d7db7"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0de79c88-bf5d-53f0-9f99-77f49766480b"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 4125}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/ef76f375faa2cd8e0447efab8e3262cc5fbe129be90442067c142929937e4e1d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c476dbd5-9a1f-56b8-b4fe-4dd239d9a8f2"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/179b31d40484addcbd3a17e723444b75855c1ef9ca3a3d8abe7e24f2d2c9e4d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c28c370a-48bd-5f55-8b1b-4b915ecbce09"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/993be8c28e6e70d025f91080d81cc87b1a67280e611460b760901c319a556727.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4d4a7efc-7dec-550c-b41c-ea5c87568f7a"}}
+{"definition": "dense_gemm_bf16_nt_n4096_k4096", "workload": {"axes": {"M": 16384}, "inputs": {"A": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/16b6f4a3ee92d9d09a750cd11bf0ea496d7855fc6479ceb6c45b07a412eb81cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "W": {"path": "blob/workloads/dense_gemm_bf16_nt_n4096_k4096/923c9fb20aab00ca09e236a1882a7d4747155aef8723da28e50ccd088b32bcce.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8324365f-527d-5871-ba3c-5bc0672278aa"}}
diff --git a/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.jsonl b/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.jsonl
deleted file mode 100644
index 65acd6b8632b7a987ff75e207bb6c3bc57a9b110..0000000000000000000000000000000000000000
--- a/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_g32.jsonl
+++ /dev/null
@@ -1,104 +0,0 @@
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 768, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "900885d5-e628-566f-9f35-d67aa7a15924"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 17664, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "88e96974-46dd-55bf-a772-be150b017d9a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 2688, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "83936513-cfbd-5365-b669-4af3541aa223"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 20736, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 14592, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdc0d189-a357-57ee-b045-c08626f15054"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b1815f1-1059-590a-87ae-21879f49482d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24576, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 24192, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c24c30ba-f971-5480-9756-4906df92201e"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 11904, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 18, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 12, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 5120, "K_groups": 160, "K_packed": 2560, "M": 6, "N": 4608}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 768, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b5374401-90e6-5268-96c8-b315c811aa64"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9502d66f-df7e-5068-96cc-e538986fec52"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 17664, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "acae606c-4112-548b-b870-cec7314a3f82"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "626d1307-1493-5593-8079-d1a58b9008e1"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 2688, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9972c26f-b34d-56a2-a043-a50684289467"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 20736, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8813329e-c415-5901-baa7-b9f423f06144"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "671e5f78-5429-5d4f-94af-24627c22937a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 14592, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "926dc757-1514-5c87-9773-99d219d9f473"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24576, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 24192, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 11904, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "22791a4a-6151-5445-806d-c3793c972208"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 18, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 12, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7713b508-dfc6-502f-be67-177b21b7c674"}}
-{"definition": "grouped_gemm_fp8_fp4_e384_g32", "workload": {"axes": {"K": 2304, "K_groups": 72, "K_packed": 1152, "M": 6, "N": 5120}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "94998979-421b-5dc0-a611-c2abe525b98d"}}
diff --git a/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.jsonl b/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..bacc6b92055ccb29c2ebc0549818a38be561704d
--- /dev/null
+++ b/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32.jsonl
@@ -0,0 +1,52 @@
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 768}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9ff70332a7877c5cdd9219c0bb58dc070d99f89d1ceab387fb2f3a4f72834ab0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c20bf0a2999b8b5a1d8941b2bb952e0317db11596b0972a3c48c56b90e459f9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "edde6f2e-b9cf-5ae4-b99c-4e313bf6199f"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4908ae1f6951afb5d038b490b09c41bb7dcb287a809107c9897d93fc21b3a056.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/83f251154e1fefb0429017f8a40462c56f2e7f02a7ae1d243ee948e77c46ef7a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "70fc1ff9-26db-5fb7-8873-070ce7537f56"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ad8fcf44c02a3ff7203d798bccf40b0ae3e147ca4f55479314fb9529efd5c3d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/177e9fe100cee2a6e026936f6623ca5ae857b338aa85bca49a45332a20a3fe16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1e43d5c-b780-543d-bd90-cf2550354d5d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5685175db75f6612ea67a6d51ac6ba5a7460394911c437c42ce35ea3b67edc09.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4059b032154ec8f1f1322902eeaa9be35cb1ed73f19924b1d7d0c61578b5910f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f11431ca-994c-55dc-a835-d8cc75635ba8"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/04c64374cd77a43c011ce61b9a8fdd6a7cdfe94c19805c84939b1d42be99e562.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6dee3f6e4e087139aa48b04add97ff7961e21de092d37430369dd2c7fab73355.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "23dbe7a6-7be5-5b9b-bcdf-8ce53441d95e"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/8a2b9d099b8e49c781a1a8b0849cf9734267f63b3e24422beaf0f069c774ed69.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4d88666879b53edb6a06ed9291756faaed907c108e8109df902c7957f1d6c5e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "998dd71e-de7d-52ce-b64b-fd65c93b2c8d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/03e4eb2d2339b2e0f2252f02ebe847fa77aa00ca2cbb6d45c5319baa21069dc2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9e835afe28e89d5f380a91fc12f86138f5d118bc5db735fc14d42d192f1b10aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "900885d5-e628-566f-9f35-d67aa7a15924"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 17664}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/825d0e41a3eb3a628137e5682674a02a19ce189e21866367502e7b29e75c2fd5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/30c07277c434aa4e2477197d44017be9410d0bf6ea5ea9b256b2857f6088bda7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "630f85b0-9ace-54b6-88a0-08bb930237f1"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5cc5d1dd1458f9474cbe84c99e35690af27c2a737c46b3015a1ea26a8626adc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e483a7d7396fceca58f88c8b282a6e8ebedd04b366f6ec90fca23b1d6e7d1f46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "88e96974-46dd-55bf-a772-be150b017d9a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f3fbe193d0aca69402b98189497f072d43eb5dfe172e4dfef75b7a863129a4bc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bbd53f31d63027e93db873c58aac26b6c8f6e99fb4b1a5d247c98f1964c2d02b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6c0c3272-75bb-5404-adda-77210a1a9ccd"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e91e61533f43fcc75fc2d7497af99074fd67574561de290272fc1a9c18df0730.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebc667df5ca9c7591014b55dc3b064d09b8fccf7c66e55b8d1af9aa9f63487a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7fcc5dc2-4220-5464-a732-5b282de66991"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd82edf8899781174e3baacfdc6656c16543d5ac72892eb87ec9e74977508b3b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4a1c99c9e311568793823bc594219382de949f85b8d5ea9d13b73afa06afedab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0ba3e03a-7163-5ba9-b206-d7bf44d614c4"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0016026c7e4f20ce677f714084c0fa68fab010ff39396db6c1465558f69ff578.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95447f3c0cec33ddfdeab22114087c724d97f6d5f921f93577e000532e2415ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "68bca134-bc2b-5e14-9ab4-68e182920db7"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/850895385a0499c54513330c59377e12be5589e88e6c85e4919c6bd386d6582c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d5115f3c2e83004d5fbab5ba331508a95a5d04e96c68c6185760d4bfc6976ba5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9d2e0dc3-9185-5307-b78c-7a8c325db34e"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 2688}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/11a1896796b50aa4a9525373259236bab3d9a24d90e1148d2af8aa6085601b53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a20d76b3e32a922820712ea71d8bed1bca819301ac45c6476f2579957911f971.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "669309b6-8328-5a7b-a2ff-dcae81dfd4f8"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4216e02c9f63cd5b7c87045b5eed2c7de4b659bbac37713e1d1bcaf62faa3c44.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/07dc6b7880e8ddabb9723ccda0ecdea8366463fad4f4ffa4786eb87c62c180c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "42076ff8-4ca8-51a7-acd6-22701e68ff25"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e164cf7d8d646ad849d625c917fe3a0a4176dae11b0d4eccb55a4e5aabcd330d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f0be0ca85fab088ba8dc352894df9dbca5e10091f7e43ab322862cb96d1f60a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "83936513-cfbd-5365-b669-4af3541aa223"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/828c246de113b567204c9ee90f94e67060a9a53d922ad0fcbe6dee103e3ba924.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0435ec64f2a826ef7eb50ad8697379609ac6897118f2c2f69e3f0ee5be50db41.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "448d2841-0915-5817-b45f-0b0accb0f09f"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c92285ad898f0ac96d68bb95d8a9482b1700ebcc3c1faacb73c696ac73aecf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd3000190c6573ca82ee419e7851bb7283bbb47ad43f88d9fde1f52f0fe6e136.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1d7e47ca-8d42-5103-a323-4baabee79fce"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/edc52f3a66e8abf7ba60ffb12bd9236adc1fd84c2503b54df9dc0d5dbfdf5ef2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d30ab8b1951f7e9a8d643ba6952c43c1a72da8424efbf5e8131a0e028e02f0d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9e4664d0-3249-543f-88b7-da51dd8a7bff"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e47d9e9ebdcac9c90bad98c036dea056361f9d537af2d3c7ad0ef2d4a0e1cbe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9c0cafbd799db06ea1575d21a1d22f5027c93ccb0f6bdd47dd53abeefba43a83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ebc17e6-ac78-51b3-99c8-183fe2190433"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7c541474d56f7a1c6a8ea395c8a2b986821e3f72cfa871671771d448e6ddd323.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/65add8307ea084a47775f0383952b7dabe2a473615e87d7271d1e932eb2ce770.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6e152904-15af-506f-a1d7-c4dc0b73f749"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 20736}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9cf82ddf2eb7a094e2246a302c0f20ee12fe2070fa1bf1f704726ecafbf3ce30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e0b2aa79d4152f748b7b91452519bd5082e7fd50bc404ac6455b06bc7f9ead3e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a53c635e-3cfe-5694-97e3-b0572332c136"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/14c32e798cfeb975e9427904423c97a0ec99968efaa58820525394c81021ced5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5ca449d22a1311a434af2a038cf88e733b76b1e7efcd923c445a1c95860ea3ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7155ad72-d765-5626-9fbd-b1c55d02ef6a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/02192ae3273a0f5df9f3a5036b159b3e1257cfc3393eace39e2da0b9e2f84342.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95ac8e0813853d83360e76e1f437c11c3a7b5a581cff023e831212a7957c63c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6a4af5a8-f47d-5751-9fe1-a984e78bbca5"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/999fa568bf1167361060f8aeee120512261e1d6be603848f1d6795575d01a888.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ff92188e4090f4bf3e86f87f5d8b0e9eef8693475d853570a7f55324f516fe46.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "025dce40-e338-52be-a06d-c702cb4f4eff"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2d6f15211875f5bfce98218d717670a86fb8e1a7b386fa9aea70cd98cdc10799.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e4cda4497363aff232b968fd7bc11b8bc91c466e4c82945846a93a4d5cf9d504.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d0761c7b-6039-526f-a122-9b9e39f94b0a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/38300850586816fda7a376c508af67588d1e069f9b9e4ad327da1b796ea931f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/e89a14508772c7149c2af3690c3053500d050bdfc93a9234d973f7a56428998c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "38b9f50d-a08b-52a3-a74c-bebc70056a55"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d26bde8c624b1343b3a74a93856417f49d2efcb1f6a4db266533eae7f2cbd0cd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7e03ebe365d9970109a761d12732601620e57b1dd8ceb1653923ebbbf1b27980.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "95953dcb-b094-5569-9867-53b56eb7fb4b"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 14592}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/89aa85b9e8cda081a4f3d491799b69c69471a085fa01782594a7e55e62e70371.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/74b938c52aca96d4f759dd8e3822e11b6d0c491b81fc161f1a1aeebe2939f140.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2718713b-be58-55b3-aeb5-e3a38cce6f96"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/712c0d336fc8ab71845bac7c54d022c841301535cc89917bd0136f1c6d4439a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fa399166a45229849717a948c7ab0ce2e055e8dc3d6bdd1eae0f9d6d255c5c0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdc0d189-a357-57ee-b045-c08626f15054"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/47bf3b3006ec8b41a85d749bb3c43606ff4142dfff4b6a15e907c4812c7e8780.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/19706542d767b94b6230f4826c2d312214b5e1797d6e7eb0721ab5f0bc7fdf9f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1034f7bf-069f-53a4-8a07-bac0fbdc3eec"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1b182a8025d8d1637d6b79ce25f7acee7c92c9df67eeef30e8a2f69fb2a1996b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/90f31c90c58c9411f0199a74431f15fd5f2277a82649124f7706dac7c4dc7fb9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5dcaee1a-f718-5532-9191-3024e0c3057c"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b2b40781b86c33e9d8518b8a2c5d23713e8d373cac2876728617ca4123863e53.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1920957033f5f90f0fb5b6f4f7e1205719ec801d233ce336020ff2c886006f49.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5e05cdb6-af1d-5a5b-8be8-0341feb4e34a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/9f12722c0f0cf9f3662c36f503413950219ed2f2e4d17c0ff0c9d1dd4b5618aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f7697717c5b5b0122ba108ea1939b6311b64ae7d04f7b09ab19741d12b5a0106.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "40955dc0-c743-56ea-9809-1954d1c880ca"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cae0a913f9c90d9fedce23b54a4a95cf0b1134ea521aeebbe4be1b1f36d51809.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/334d750324273693f22feb0ee986e85dfa8cadf44012c1f73ebc5b39a002c878.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "888761a7-d8c0-5d0a-9e2d-e206473eb438"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bdb8aeb6a85c032dfef4bfad0c100e7a4c62eded8b5323a8daf72c64d454aa17.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/10518c23fca200233372abdb306e6a500145ebd7a2e15c8ea84341d13a7d3bc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b1815f1-1059-590a-87ae-21879f49482d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/dfa1946b05833626412e52e9c8be13d0123cea1dbd41565278c347025ee805db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/063a7d2f2584151620afc4a53b92d49e5a2e93872b84827d3d9464a368bedd83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "75452890-0209-5264-bb51-e75d6c5b75c0"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 24192}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4df7da740cc5a30d1f41450e905bf58d1399198cac9ef536167d7138cb155c36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2ec02a62ebfffbcc1f970de736fb77b4839221757e17c6e47e061a43eb08844e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c24c30ba-f971-5480-9756-4906df92201e"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 11904}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2669496c41bb97e479c4fac6a16ea94e485d1c31f44843a1317e74a6c25c1ad6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/33dcecebcc3f083d136c30a757b24b3e9106b33481850e74e9b81feb5f0b4972.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7e7da669-03fe-5ee3-a1cf-b496e7614a56"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/235dfe4aa80da51b27248edc8faf269128572f3772117e7b76c709a431b6138e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/24f7411bb043a870d90d9df3e302831df3328cf4e086405956d7c208498badad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b4fecb00-3198-5a46-b718-a14bc4361694"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4bfb30d6f6d67c9e82e3a992a40d8a8c1843042b8c65f4501873c7aedff523f3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/2e69d9189c24f65506c9a3196cb263056302e902cfa39815f27efb9c050b6748.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d4b15562-fea6-5caf-ba58-a6ede4195c3f"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d80add032aa72320f20a48fd8a01ce25de4439d90b9817bba2f1d8eefd0df06b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/98f4a13ccbd2cbd5ca032923ec5f0141ee3de4dc405564f5b8cd3aca45bdcd25.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3dfd91e1-59a0-5a61-89b3-3434003ad831"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/23a4c2128af4eebecce6a7aa409efd6affea2cddab9b6b9e2ecf881bd57265ce.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/903cf930ebc0d356d82b326dec1f398967773549dbe014f7318eb6415d1c5236.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "759f9ddc-17f5-5cdf-b9e6-d7f43e5ff965"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/bb34049b91402e230e37c34be67639a16cba01524f6998cc4f0a5d6d1854bdf9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/20f97d6135c117fb8a6fc3af83a2a11bc72183509855964a938e2a4a8ac48487.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bd5539cb-9883-5d6b-b935-0fc2eabb4551"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/332bb1154a0ba00506a6f41191f16151a6f43dee4dfa69a45f132a0ad3b1608f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/96013afb76bde35b43e1c50fd6b7e3f947be3acd6dc683c28d7cd37f8f15d491.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d500197c-537f-5a7d-b132-5dc3b4651d12"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a17ee7ec58568195be9a09722f4afe00f7068a7b4bcc232f2c5c348dc7b908c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/28f98c488a965077c2e15ca919bb744a168d6c678dc94c76c323d687dbed588e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fc247ae3-7332-52cb-9e37-11d6417daf81"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/7b788c6e30d174bc3ed64a262265fac0ce50d85a1eafed555eb06bc7c063a7bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/b6153e9116ed1fc9a62ada00d74dc4c2bee1ae4b7097febd6c26b78c21079212.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "54cc9aa4-83ff-56e0-830b-a9634ca9a54c"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/4432263927b7b70ae52b1c6bb6e104af5446e67bea7cde3de973322241d8f680.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95bd7a9a1d67c013dcbc260df7a7492793b89d3e5a056d994e433726a858199a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6e152962-f1bf-5b6e-aaa2-e0a40cebc869"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/aed9f8a158cbb2caf11bffbe3b82a3b725cf444657daab3c3ca7f7ef71eb8c5a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/cdcefbf33a8be6cb4729284ebd5ffd6e63bb38c207827e7b7e189d2551330cf2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8f39e29-79a8-5f9a-9e6e-931a67868ed8"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/3842012205a0d04ae566421788844fed2e9de0a6c55b31b4feb30a3d6a6983ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/fd550fe2481cb44d2fb7c6850ceb09aa951716a87c3f8711925af9958c87c5d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "16d8d09a-daae-5853-834f-2875fb85a5ae"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n4608_k5120_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/588bd4b783134088351aa6cde2aea84b33d71b8f0c67ae2fcae463eca993ffe7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/975caf10027419cbc722e0bc7cdec951b5e22dfadd35b08e0423345fde4fd8d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/0d224ca7c5bd6c7101caf8a13f62eb3c156b557081308c53dac274b03b0bf5ec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n4608_k5120_g32/981e70c27bc1fd36b6a67b6a4faa602cf33fc6e81a3771bf46d1e3d087b42e64.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8ebe3e49-d085-5194-b8ef-3b26f685f3d0"}}
diff --git a/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.jsonl b/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5e2ce6c7d48acc7ca9ed4529c76dfff490a686a6
--- /dev/null
+++ b/workloads/grouped_gemm/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32.jsonl
@@ -0,0 +1,52 @@
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 768}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/857f64d4418d953d6877d07acf88fa894c928efa113b2feaa3fa8248f9140b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3893e60a258402e35ad4bfc4c1bc15e9ba0fb0f4137c32579999cc61157e542d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/33a1946e1957dbd2ddae6de8d9a907b350a41e649f842ddfc23959ae3bead5ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "59c73c1c-f9e6-54f7-b15f-f6384e355503"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22a28eb898ea3dfa0bf239494267e66c87fc94812f04e8a7a0d610701fe12cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bca5d63c1f35ffb5ba99bab4119222ceff1b622c2ac4fe078f77e1c22ff98308.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/947805815580548c033d096d8a652d1db49242420e3d822386b9bb1305f3e1ee.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b5374401-90e6-5268-96c8-b315c811aa64"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18db72d830457629348005e689eb021c57f16fd050f10a6a45baaa18f12a0fa9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0ae8f308e6e9ebbfe3b2f6ce65829a1ac8a6317a8b369e24d86287aa1f7b4246.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/18a43cfff1fa2e0b886481baa4829295015da54c5c0b511fba4e2635fb12cdd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8d1a3998-f38b-512d-aaf3-4d76b87a42f0"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c698d3af06df92f966b8e21331bb816980582b786e25334ba08c683d4f6e184.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b72f8f98c9e1d7223498b5ee97a227f87e041a55c036175bd6200080307a91c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/59a77b116773f6fc138b1aa4d563e820f0e02a2f1870edd86b2d92653ca2fdff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "658b6b66-5595-55c4-9ca4-fb508696cfbd"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c2699d256cd9c38b55eacd8215f4c285e62b18e6b997017753123ad9595f20c2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5736723f8f7573eec09f9ae55a6950b2053f0b606df204c74068cbf556914d16.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f15ddb1b2cac6cc8a9fbc8401c10e7794f384c258feea0a99ca2b802e44f9389.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9502d66f-df7e-5068-96cc-e538986fec52"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5eb42e4c15bb7715c9cf0e4c0acdd4fb1c5f9cd82d03966a390489dfeea5dbac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ca752c899d5b47d5c2a523e0b2cb6def7cb0aa7ef0a0a77dbb014e5778d753f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/477d725bf1a1d86d132fca5f7ec15e673b075ead9f327aa7616e78c22f4969db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "87441003-b95b-526d-b0f7-9e90148ebee5"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1015d1e03b1979fd5a8a9338dc308fb893568d27745cbe10e994a7c93241862.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78f519800060bd1e54010376759969600f1e748ba034c7b18f04c8979b889c19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c69b000d14a0e16ccffb3788d1dff9ceaa4093acd95b7aab963943e1bc22ec99.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f94c4eb5-7e50-5719-a440-040fa6dc1bc4"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 17664}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b629c4d77e73416eff6a6ccfda0fc9b1b1a385ec4e67c23eb5b8b74026f94a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ba9691e351bbe2ce4de82980258423a1a1fc7d0d36462a74ad4cd298ee2bb432.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6afa61c0f8178f3feccf50e168c26b6e6d80ad4abd584c299cc454317fc022ac.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "acae606c-4112-548b-b870-cec7314a3f82"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d97c8a4cf52a2f83c2b369f3b5d80f3ce2d828095429dcf20a21afd4c33b8a2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1e3b97a6ccb1079a2ee5289b96d5276bf850f20bc3b5066fa72ea35394d2c2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5e77586c2c41cc2cc66f194195688ef909b675a4b3869ed05d33da5879fcf9db.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a83055fa-d10a-58ed-adb6-de682564c8cc"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/be83e022bcbd243793fccd57794cfa281028e3ff5fd1d199bf4ed94adc094560.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f81f13bbf96d8a31e7d9583d47a541be5e39ade5e25885005067d4c725c825bb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/31144e9775e7ac9d887cc7cc0464697b44b5b623a5fb57e77265f5ac80389348.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0d2e28d9-5ff0-5df1-baa6-ce422d564244"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/0a5b496d72a70c0460bcde6fd91c5aeea80522af37ec45fbe1efb6cb30e38134.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6cf4fe80e6d429c035efd4b28c964abda0614051e2a4d4413573ea1b0fba7ea.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/48ffc8732ddaabba24a1f6867a0d1c6921a78fdf834136bc8fffa87beddfe5ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c63b7111-d931-5b58-b45f-9a4e05244eac"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf61ef3668506b39dc3c4d1918e197bc086dfbd98fffe54d65b7dfc9baf2a34b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/539c187b52e2c54c146b0ba07d7cef8a2038e7efb78205422ec695d7399ecb18.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1ec5378f8c41fc583c7fa135f1ddd256854a5d27774c14dcd0c60c44b2edb70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1b416d6e-080a-5179-913b-8aa5707cc3f7"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/03e08ace15972ca7975425a1337eb2cf5d91fc9d6d04b491027832131a87ad8b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6291bbc9b515cc87b0284fabb5331a3d1285342b439b9fa562666cd63234e316.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/75d6c5893fdda3ac355ceb4b37344c9a2e2247848cc639e3bcc369f71ce08378.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "626d1307-1493-5593-8079-d1a58b9008e1"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/22414ef8fb448e1b5bc2a535c35ec263fdf1491b65a37647e2d78201c8934279.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/13dff877f0f135697dfaf881a741684b9b079fcebc812f13fcda821a165d008d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5bac813b-88eb-56c7-b06e-312a7ef55e26"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 2688}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f84c58b773b60ffdfd92d1e372873beb96fb0bcf2fb2bcfb95a81ba254633da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f43ca96f346e455a026a63569194dd9972e90f4530eb3705c714dc89908b91ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4454ca00aba8bd2afe2bd8eff46d8bda14bdc496f0ed62b989a93e530d251797.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9972c26f-b34d-56a2-a043-a50684289467"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9be6f34083e5451d6e670655aeb80e02e18f2d2118a592498acbff3ef49891c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6e3fabb37ca750896b0cc5033097304b09d53948438a1ae49346acfa3c4e4843.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "94166d72-e40d-5565-b9b2-db03047bd74f"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/720b96bf6c5705cda589705e48c0108fe10771704ba05df43f5b96704ecbb40d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5ac69d5590dea61129b0c766cccc8b78d26dce2d00fd632368a71e69e0f5125f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b93c8169a3a108470ad7c340e3cd212d867f93542cebce8fc107ee1d33a22b48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "73b67c4d-01a7-5566-ae5e-07adfad25dc2"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6b1591ec43bddd00b3692084982567f7e310e7bc4608a076a2c1828be15c4a40.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9fcd55300fe288c8c6d84cfd884c7c462f84ee9ff91c1ae1c3393e2d050e8711.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b9056eec63e8df0d79c5a409b4092ac6a6206af2d852cd93eb973d60dc325c15.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c5e9467e-5252-5532-aaa6-28a4e74a6789"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7e81a50d3d1c993b6a9c2496476e38e4bb54e583bcb5fd80d2d0404edd6b5c23.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/638abc853d7d0874a371896c4805279409f9ecf7e98fefd0329d2bd50eb8fa11.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/141903c0594d67f13c5e8f8457173f91507dafdb09f54e19ace3cfa323c8842d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3ba9cb47-0488-53a7-ac2c-21920bac8a55"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/78c09b190dd8669493d13df5b9823022e74fb76f273babcc9c3fb01a53f12b70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2d182285ee3489f7f0894059a7f8cbd53b9fede4e7cdb2d37de2fc8e6dba3de6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/55933848d3d9e5f5bfa54ad182e3d6fa8793215f82d7fdbd7434a1c5483e2eca.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a77e7e0-79d7-5650-828d-2739836e20f1"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/db61b44b42be64e0f49de45ffc196f42db2055070d3a2db44cbdc8718b76e7b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/767a537bad4351fcde3a795b766aa75843a4aeedb1e2e7e24804654768ba4551.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/913846b183be93d0d7cacc366fc2d6f0cf96d8affd4ad360093f40a425f0e896.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7d1d7a29-8b86-5580-b237-9e5194ca9a6a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1a3cd3efa19948cd98bcd0fc525635460c092c13368bfe84d71558dd6a9b2c31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1804c9c958f900854db1dfc06b7ded4fab76c18e5e86bca0bf7fa39e81ce3fec.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/63af5b901dd27d9ae4143e3e7c67c4c6e8536f60e972aed631fc9ab6ef027b45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6405726-c01d-5e0f-b77c-fc20a20c93ab"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 20736}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9ba53d8e06e54d5d05a6c109122334bf03d0320c58ccf86cc8fa80a4a24038d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/15b6c0c5672a31aa8ba2a7fdb857039b73f4f85b8faa06f5d0f72725ca0aa458.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f1954c64e28c4ebea6e1e8b30dad4e652e6daefbcf600858d341d0b40f3debad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "93be97fb-e453-5751-994c-2f88cbf6ff30"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da195b2e070083b268d098a89766d9b66e2664f0d36e776d155817fe4ba46d0a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/12f2ce79e7c2a6f49630ece1b112dc0101453eea6353f3ed9912196c2c51fd37.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5f69292fcb1df2d132726d03fb4c431b5382099e167e8ffe1692f1445d715bbb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "819e56ce-0e9f-5d80-ba29-2b4b66f02f1d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/276ad008e42e4d1dfd04c7e4386d8eff73ff1621ea9dfda045cdf5b8901bb98c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4e7fcf861ad3571fb12624734ad8fd45f106af437e4668ac883f1363b3e40cd2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/98442de7d104a7fac184428658992d1e6b36af061d824c1fcea106e7fabf2785.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1ef8324f-96bd-5a0f-bb09-6754b31ab356"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1406fa751c4c798dd67332e04bd998a3160696efa8c8fa5f0ccfca6d56517591.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b50529399755e8ffd80e9c26ba7555f2bfb1c1c5651463d3d089f81e3da1897.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/aab68587aa03baee2cb805ffc0c7f4bd899eab305a4e96658219640d0964a540.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8813329e-c415-5901-baa7-b9f423f06144"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5c27bd8b33f68433870d3f5b1f802458138aca283cf58502cc235998922fe149.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ebb8267e7a392455b2a64e8838873120140e6bfa67974c24539ed677dca02699.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4822f00525cdf65a55bf51e748b1ce6778cea06b7a93886d9620436a543f5c04.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b6c96f86-5c08-5382-ad66-482a71b55b13"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/41df1fc718b34bd8b45f71c14dec992b223bfd87ab33aff5b82fc69fcddf7532.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/730238a8ca74119c909c44d458a19589d818f02e549f0fc055c351e48a113c78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3553ec05f4fdbb862f1879b7ccdc9762776447e5376c7a22aa0528d78daf19a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0f147c8-68cd-585e-a737-a1da8a53662c"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/407286f47721c1249d6f009315dd0a2ddfd51949124c00cf90fc3fd873d83a19.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9f4cb3047046324c0c4a3522e575db43961739bbbdec13862b23a54031faa052.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e6c59f150e160cbd562305e0c424f0df147050f94abee9c718d88e35ac8aa4e5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "671e5f78-5429-5d4f-94af-24627c22937a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 14592}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/72a49ef7f85c091e9641657da2e5efce0a8cc89ec96b9a54e043af3cf960da2c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d9ffaa74a051d10527b6b1e3cab1087558876c0ff285d742a1fe7efc9af32f24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f4323dbaffd760055d7df16f06dd102b48ef3b9787a41cbb4f84054162196e9a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "852cde60-0529-5e42-83ce-0ebfe07a7c0e"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9c39cfa1e65f04fd6d3ba5d7af841fe98d4827db388a24f8056358c0ae23450c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/acc5cec070cfee9b0267d78fc2595d1ff7fca1c6c003c370b005fb55b20a3bb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7c679dffa187d0a9e155cedd08fba8198a39dafb02ca01ae891fccb5bf3c0b1a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3ff114c5-d974-5bf3-8502-c9c346627f67"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a43e00bc912ddcc7541b2943aa3c1db8cd14a77de7b7fd3135fed3c4c5093dc4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/9199a8a98140d3d6b6bb767f927f493ef4fe3a6efd91e1ed75a36fdda5cc4a97.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7333c0693f03361fe546a431adb8e4515b2f83b53b70d1ce1c39c573097c771a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "926dc757-1514-5c87-9773-99d219d9f473"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3c6e30f66b2a2256ce81eb2318080902ed1220f9237a27e925f489850c4cf63b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c5f7532c13b2cea01f5022db807452409b471dfb5cd38eaad2fdbbda83f40920.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/8bd7688de335250122148181b751abb26b6aaf08c14cda832b4982488ecc313a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6aff8387-0383-52da-8d07-d0c10ebf2d4d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6cfb0bb041b2fc3d72e28a7bb51b5e4b27f5bc6f272f61b8b6c0907451e55034.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/92983f38eb15b638e06d044d7d9afd6246ff2272ae8f2964b17454c913a73d9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7953f22de67b138f8703e7bb101fa477e125bd0e82664de9c4d73a6082dc83d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3e554d41-bfbe-56cc-a05b-179acb8a60f4"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/2eb4b8da4ae11f20e7be82ca79e5f115fc0dcb32554bb08f77cb37fe33b2e073.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1047a8ed4ef4730b956bfccc40a02b21973c6631b32d65257c33884e776676ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/20e10e6596b618fa33b64b69b100c904645b10b886a5c6e323ee2b40837f7bb6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7f2b006d-0a89-5e93-abd0-f8563c06f8c9"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b6bd272a690afa6ba778d20ff5cab7ab11a962b49bf2510164e300edabe62a29.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/35cefd857ff6c2a69a7949cf22f909d0f7b1f7a15b46abf43271e4077c05f8ab.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a5920ee94d671e9aed9d22e1aea1ccf8c225d5203914defd3c07c35902b45af2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0797a69d-b39b-590d-a7bc-2e7896a70793"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a78280171508b8809447f5cd6dca8d1a03d34d2a89a43ef550c8fd598cb78715.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/3939467ad831cc9c198d6f8ed6cecfaf119a4388cc67f4cbe1cc8fa97ab27957.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/86b013dc73c7aad95abb4edbbf6032116f36a1b18db44d191e97731b653b3dc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b878a86b-c444-58fd-88fb-935d7afcaff4"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 24576}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56a51325286e3c5ef334a1c39832343c64a7370bfcc63282d9fe7f5239a90873.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f381be4e76b45aa035e1eb1658030943befe91717a3718b3996b1ae94acd40b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c7bd007664de0ce970fc125a8b3312c3f5ca100adc963f92bbbd7dfec1333aaf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b1fb4669-1248-56ff-978d-f610aa47ea2a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 24192}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/145d8e4100d107fa596e9c155e5c8cf59334744314ed4a80eef628a0ccbbe812.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/cd011293700009c3a8d8564ab7fba9026dc6c34aa7bb39c406f383315aa499ff.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7b87f8b9483362767b5c1959a616564919b7fffa7bef55e36e7d1da5486a8d0c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "a51a750b-1fee-5798-ac8c-6254b2d4d4f8"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 11904}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ad856a43c88d1c1ffeb3e2b39551da016d5d0d07b4dbdd6caa35fbb7afe7e750.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/05f465947bb92d073e73596c036a33ec185b914bd17dccd18a549401aeca5cfe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/043edc9652b3087f07e2daf733b4a76f8f9dcee3a80401393ac441cda2aec7b6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7db1bb57-9034-57c0-95b3-2a45bafb1409"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bf9e4de9287221b25db34df5d9a3fa45ff3234f2a15a134d60c00563e4fa7d12.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/70ec79da6ccc431841a6f6081a854f63afc00400269d9a5dfee1d45a8cdcf1c8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/74137e55ab3a13a9e15d311139c772a284bf846529ec58f0f2f559c45b6e7b9c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "69b2486f-f566-5982-ac2e-8c800ec1dee5"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/57c3b993254615dc3753de4c31e5a900b332fc7860d9d50d2227574d75e75775.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/bc32f6b55ae418c8dadbe85d19c1da99874d37d7f09feaa58a4cd9e1df7b5b7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/7262851ce605c818607c68731dd76b6e2e41c65cae74afec9c81882db7a15d1c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "22791a4a-6151-5445-806d-c3793c972208"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1020fbae5934e3b5767c4c26a5eddc8ebcd2fb0f08e6bf147b5e9839edf1ed6d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/5212e86f398137e6b53ba20a8d8c8cdd98935b68be0aea47985227b70fcd20a9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e7a6ec243a31b759b1b0eb7e4c5c8ea94ac146cecc25694d97e3022f78d18b52.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6d903c42-3165-5ccc-a2c3-3ddfc8a7ed4a"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 18}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e29214a7ecab153a7b3b476b24faee5b092c177544c41f64a7e22746a7a88478.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee30db43c33e1bdeb047a2ba5f471663eb56c937e7faca9cd6de024cca7255c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/10796b6ccdcbb26d92ea58f68478a69b490ce6a78d4208ccd2183233712e482e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "82fbf85f-f1f5-58fe-a053-583a68c3717d"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/ee75b48f06ee451b08da57bcddd8f5091b654ab9d24d07a19394243be0346e70.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/4779d5f4b47b422185371d22b376491c0369983af5ca249c973a0341175c7376.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/56f44211b5f23807888e5de5976313274a7fcfeff3648ba09b376e7a4a339e02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9f6a3944-09d7-52ca-abb8-4b01960f3577"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/625919ef826f28bb7ff134af77ecde7020686557ef5aefecbbdfa4aed1889ac8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/174b64eedd36725a64a34c1b80912a97b1937c28cce31d910a7d1e06d9ebae31.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a8a455e716a9fa2e85169daa23e197724064bbe9a089dfee94fa5948a85604f2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f2fe6eb3-e86c-5e2f-beb3-1de5680a8044"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/b65b3ffd73529af4eaec7145102392c5ed0ef6509d86e80a8e92cac74fd48d10.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a1c6941e9a4fc6ca61f4babdccb759bafe8ce5181ade8e5ce7ce91cc27b0d19b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e0532f1870cd59c9a5a39eae07777d5814628da84694c77b393b6195b01cb1c7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a62ee81-50f8-526e-a53b-f75d3f5a03d5"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 12}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/652e985e8fbbb61dcac00c47add4f598898ef0558dc9538854e49119464a0198.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/534e133fa871024db798fb642f7aa368e7ca30e30cd76a8e6a8101331bd22146.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d934de2fc57ee2e7cf47e1c54da56839753701b78460190f8b9a4a5d86d7c473.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0c7231d6-7f41-572a-ac47-eceb974cad41"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6d873456746af5cbe33659f830e4832b1a4dced1252a3127d6d7104dd78c3eb4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/905443e8e442c0e7cae41550c12912d592b8e267470194fdb6ef69009e9938a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/38a175bab60010ddd5683abb9c9174c43fe4b050e9d274359e910eda7cd47e9e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "925a836d-0688-53f3-bcd8-c8b1a67135dc"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/da0e6babbf39b417c9d9e3ecf3f845f46437043f498492243ecbfc40f634018e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/6a5f0620317472f0fb5486a071e9708965725c5902d83a839b9e2b9a4a6c51b7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/f958b297f44284e9d5c1e836adaf8bb7ba0ddd731733add95b3d7f32c41886c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8436b1ad-0462-5d00-a84f-b89036d2dfc8"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/960d64c8991120df109efdacc166bf44cf869969ff3eb9d21cd46bab1f587863.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/d4090282537a301b4e72f09e1765b277e9863837a3db4756166ec40e55ba2ab1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/c08ec486ac76997f701491ffa2c017ad93ac6c27bf83e3855aa9f54040b5e192.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7713b508-dfc6-502f-be67-177b21b7c674"}}
+{"definition": "grouped_gemm_fp8_fp4_e384_n5120_k2304_g32", "workload": {"axes": {"M": 6}, "inputs": {"Aq": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/80eb32992544b0802ef7452c4985b281613d2565f35eea85f18d62956cb6338e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Wq_packed": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/a7ee38a798fa3db1e8deece0fa13863bb6a035cc546b0c64521789172ba1b833.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "expert_id": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/95d17682dbf9921330c838a4681a70c7889f4e8e6725972e09660671a71bb5d1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_A": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/1b462986e6ff0dc12c40b0cf260f37bb1c1fa5c5c2b062e4d58849e94318a767.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "scale_W": {"path": "blob/workloads/grouped_gemm_fp8_fp4_e384_n5120_k2304_g32/e9726fdb902913384a52fce7eb3ba2cc39aa9f284b7f8abb9f50d05de83b897f.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "94998979-421b-5dc0-a611-c2abe525b98d"}}
diff --git a/workloads/neighborhood_attention/neighborhood_attention_3d_d64.jsonl b/workloads/neighborhood_attention/neighborhood_attention_3d_d64.jsonl
deleted file mode 100644
index 375b6d2238b650d8d0b6491edc32d1e980b60ea3..0000000000000000000000000000000000000000
--- a/workloads/neighborhood_attention/neighborhood_attention_3d_d64.jsonl
+++ /dev/null
@@ -1,45 +0,0 @@
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 32, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 32, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "503c803e-a524-562c-a980-daf2bad42368"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 32, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 32, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 32, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 32, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "961d1949-75f1-55a6-9950-7c470869589d"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 32, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 32, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 32, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 16, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 16, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 16, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 16, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 16, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 16, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 16, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 16, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 16, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 7}, "Ky": {"type": "scalar", "value": 7}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 8, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 8, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 8, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 3}, "Kx": {"type": "scalar", "value": 5}, "Ky": {"type": "scalar", "value": 5}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 4, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 4, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 1, "H": 4, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 4, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 4, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 2, "H": 4, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 4, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 4, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6da01593-e133-55fe-8eb8-69be7217be58"}}
-{"definition": "neighborhood_attention_3d_d64", "workload": {"axes": {"B": 4, "H": 4, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_d64/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Kt": {"type": "scalar", "value": 11}, "Kx": {"type": "scalar", "value": 11}, "Ky": {"type": "scalar", "value": 11}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_d64/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_d64/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a"}}
diff --git a/workloads/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.jsonl b/workloads/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c4fe43c7ed2595b730961f14b25c18b4e4e6cc58
--- /dev/null
+++ b/workloads/neighborhood_attention/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7.jsonl
@@ -0,0 +1,9 @@
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/26b665c994d49609bc867753321d657ac5a260e14d31454745589cf0d17641ae.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0a1ff829c941cc8bb53093bf1e2d29e0eb5b311bb13c2d03cc3e6dc4dcf487f1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/107efd2f011f75775ccd8a40a66dea0e377ef5b5f812a72d187fc435e4021cfa.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "572aaf07-1f53-5279-b4d3-b8cc9ae232c8"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/7bd56559fc5e978aa270904f320703994dc503b4ee4e9817173a810076578d72.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/44604b0e762b86f8bfccdd41820545696185321262bc199d462109a0745c99d9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/43dd7de2790b5dac4a91b2017e44b6af49c81188464c82e724349f41963cf2c3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ba0a26ae-0688-5564-a415-b40944dc1d2c"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/b3335557da5031e0a1215299d1c07b80d559635b4643b0338e0c857fdd5fe3c0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/c2626b6553a852b4a1b83ef3ae0c7e099b85ec265d855b7dcec31f15c838d56b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/1203a67b70e0a633776f5c9ab78670c9cece7fbdabdc8e1e95d289fe3dc35ace.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5b35db05-d838-5b18-9415-11dcf35d9cff"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/56d24e802a04f5cc5c707906e017f754efa5ce310f367a469a23370cc2c2be0d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/5a497d2e4ade347270ea46ddf8066d43b1c1c60508b39c206a82ea9e0d40cbb2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6feb47cf06784819e57d3cab5586f2307a98e92a8ed7d2771d08c75521dfba96.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d6c3cd98-bcbb-5014-96a6-e228b4157aab"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3db6db4b9d85649bd52e6fae284c5643d4e32d221e85fd7cfa283b134123536.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/3401dd3cb301f5ccfe8a4301cd86f4b5032dcdbfa23f73872495ca8c4854f3cc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/59c241dd27be2de0a14484cdf381491da3921780d2bc307f5e3af3342632daf8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "06524fdc-1583-57f2-80e2-fdbbe82c11cf"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/a41ac6cfc761c942646519dd9ab6e43b96368db9135ca720836fb576447e71ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/6ee31db9204de3ff2e9ec059d25ea643b476603e00dbf5e61cd2a5558738a670.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/dc3cb6cce1a1cc483d0c1f668b28a5e0340d9c2547fd797f277721248c79cf10.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f8bfe021-10d7-522f-9a13-fec1520fd1a9"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 9, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/944e6730ba5fc1516b6fcbad2a73fe41a6e2024fccde441c7848bc8ce9d465f9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/9a9f46c2613c8613d857c3259a39f0e4b2a08158d05e1c792b6f9ebe22b7c851.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/855c3e18699e3de87907f46a68eeda959868cf4696cbfc2f67420bfba37a11d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7a7c903b-01cd-5e52-a7ad-610af7f4b4d4"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 9, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/936aa8594ad39f3737de83430baf64142923448ab8dc838573d11b21b0c2e1a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/841d7a4653426bb8a8496e1c42a67b0d8c3c91b23059d570b8787a122fdab8f4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/0d048ba4d3f1d0676be0bae28e663b601e03734852774f4d197df8f6f39b4b6c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "328bbc42-0b45-5dec-bb11-0a53ec0f4ea5"}}
+{"definition": "neighborhood_attention_3d_h16_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 18, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/d3a378b89cdeb61676327fe806c3e73f6556a370e2333239867fa0258655c6f0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/119b84962f1770ad0689d2c7c0a1a94e1b91b201016e1f7bb9baffd7f03f5580.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h16_d64_kt3_ky7_kx7/986fbd99856f4f222951b9df6200bb527ef8a92c2c08bac39e8e55fc530a0cea.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e8f46b73-bd40-55e5-96d0-32dc0199b4f1"}}
diff --git a/workloads/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.jsonl b/workloads/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d9337f86950b6c300f35c7efcd76ca1a5ad61401
--- /dev/null
+++ b/workloads/neighborhood_attention/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7.jsonl
@@ -0,0 +1,9 @@
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8662c00f02a16260fcf407b6a006c0346df24407d7d7d5fc73fa45d245abe269.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/feae3f6015c06d9e76737fccdf9817a73f0ef38cb6597eb07b5dcc112cd3a008.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/978e1dabdb9029f6b8c8e0e96538c794a4c50bea043e2612c39f29aadc9b5c42.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d30833f7-16da-581b-84f4-66007ed77ae3"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/e5a1b72c52134abd5278d316f4662895c4b2190addfb7206347f3ffb7e1b0d13.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/654805034928dfec54085c76d03083c6e9b9a2895cb24bbdea93f37e1d5192ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5acc46a2c1cea1d105a7beccf8f072db52e20a17cf02f33d8852d5294ca1fbb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "503c803e-a524-562c-a980-daf2bad42368"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 1, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/04cf4a055b7dcc041a8002511fcb7931a2f92eb4f3e68fc9627f9fe0b0871a05.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/3c4b0410db0c56a59be8499c10c070ef068c8e8916970df75a6d5a9f72c72765.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/13059b1c7b565e046027f6b20bac08593f578d6ed4e8f4975a09811105d346db.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9ce70f4-0314-530a-a4d3-cb359744d16b"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/bbef36fea469a217ea7a9a6115a92027609cda3b12511404b4d6885a194cf4e1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/84422de16b5c2c6b58c33cd9e467a664214636ffefc268f40b9d87266d1d5817.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d7b819b8707a2d3e199dcd2d5f4030801a30e5d3a16a79258a26431864fcd490.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "01fc8439-9178-5ae6-b4a9-8a81d737d807"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/0ed4110e5ca8457f0c6fb7622412143ff827a6dff53f8a8c74d6af8101bff1da.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/f7ebbbf873297e379fce6013c9c393f939a29951bd01ebda9f0931f9590f2518.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/edba00b8a7a4786c2f2f5ca6955dcebf2aceadc9bc8fb8545df8d8a2956a97f8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3b569bfb-00e7-594e-a79b-87cb68e3a016"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 2, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/59463dd085e0328a4a4a6dd8812c52781143c1bf280ae3d80a0a1255c0862a61.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/95a5230811860101e5d845349142525b4218f48d6efdb992709b15c9f6284416.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/8599bafb2c6a69d92c64e40755a38716b3765f6f3a07ac5b4496c92b1e3f5d1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "961d1949-75f1-55a6-9950-7c470869589d"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 9, "X": 24, "Y": 14}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/d3a7f58c1d20b33de50ec550cbdd69b415e3feddf65aa11c18ceacc6f125fdbf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/7777686487ffcefbfce4055165dafcc2b0dcca914d43d4c5e0774cbb8667c65f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/9dc97f305c41184b37cc19aa1c0207c0cdad324fb391eb89be08025b47f6ddde.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2882e24d-2dc3-5644-8e69-8c22c2b13dda"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 9, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/32432dcc40ce4885b0f27a523052b0ced127b717d95c732e4ed851dfd02fb08a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/dfcd3fc6dba0b3e4d191f25abc08a269818688ea0f57b7457f1de965f9db42aa.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/b1f53965fcc19e9e7cebfc4eed4bc63cb205b84d9862f80968d531b785f87ca2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2f58848f-7e64-55b7-ad2d-5de5d12d6373"}}
+{"definition": "neighborhood_attention_3d_h32_d64_kt3_ky7_kx7", "workload": {"axes": {"B": 4, "T": 18, "X": 16, "Y": 10}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/781bd9fdc34a061ac58631c43fc5fa344509b32210e78ed4e7670aa8db2d1d45.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/6bafc06a982fc373a0988725f771b575143faa94420d54c9233a9d48a1b882a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h32_d64_kt3_ky7_kx7/5ab2b690a41d57e38adb6c5c83000f0c04a6cbbd47f445c48a56f37ce34ce90c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "867a9c5a-089e-5005-bb09-db1354c3e298"}}
diff --git a/workloads/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.jsonl b/workloads/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..561fee7a5970a487c609459a769733d6e2cb5cc8
--- /dev/null
+++ b/workloads/neighborhood_attention/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11.jsonl
@@ -0,0 +1,9 @@
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 1, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b31a3dbc735beba8826eca30dbe7450b8ab308dbe177830f4a81789586e8a2c6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/458ae2ede2919107934d55901acacbdc30cd9af487a9ccab8e3a4c20f4b6a242.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d88d08f41ed752cd8fb3478e2cc8e875806e201eb55d6f90c575751377262948.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "bb8a6ae2-035a-51c1-b6a2-2aff689b88f1"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 1, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/8fa1e03ae9a63bb78dcb510f529ad0a1f8ea9cc59f5468812339c14cfdb47e78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/c5f2c518c48deeadc9f659b090eff1d4f135126758b7c76d205e59a3d5b7c334.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/48e1ebaabc3f7e89999f4fd6fcdcba5d25e5774fab5c34caa01db3eb55940663.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "4bc37940-bacd-56bd-b86c-51279edc24de"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 1, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/624bac65242de2042b6aa7f9cad323b40b0032d4b1a178ce770bf49c9ea4aebb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/216d7ac42316c793a948f9e4050e0e52aeba9b04875fdc8a90d3ac3ec74e15f7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/af3a860069f9361ff052c73775c0026c0ccf6c17926fdcafcf608782710e5a90.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0fe22fd-bde6-5d95-b486-a1233f5342de"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 2, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/31733875763ccff802ff8297b98d2a860309aaaf7694641e3ba07181ba0c7a02.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d7e89d6f5670dcaecef562fc49a02ad1077424b00e4bb13bd0c74aa48342061c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6c53f95f544dd55b457bc49a4ca4c398e8cf97b64f06efb909bf89fc22f15f9b.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7df22f0a-44c5-5824-a3c7-8789320c1fcc"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 2, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7b6acdd35badd65471d804dd84836fa6a48288ca9cff50411539e7a5f3a724d6.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/7707e9b1266fd65f7f7a01449f14bc100fd2a43851f71569fdc23e034c30d291.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/a8eb2210721c35b446a8bb561e5b69152c53da635b84df565d8dd923747cb1db.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3aacba35-3f6d-5c90-8c63-a78396ea8d44"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 2, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/21061575d3192384b4801b15622838c8a2764b85e803e56dfc7b6233e787046d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/607487c570657e5b626f4816f486e08673f49ea88cd460722832187cf5a13f30.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/06e6199dfe7fed1508ec5da169685e1684329bd426553eded4a1d20d6188706e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6512849-69e2-55f5-8e33-e0f8e148388e"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 4, "T": 49, "X": 192, "Y": 112}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/b7e80cf2646e95e5c5156900bb5be000371ec0833f95a0f6ebe330830340862b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/1203e2afb64d04b790fd36e94135d9a42fa7e9f8b4875930e08fb30deb133ce3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/3a1760ba441afea707b603c6b368dfa019bddb5d6ff4d031acd278810ce7dd04.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d115c6f3-2ee4-5393-affc-0f41a03e63aa"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 4, "T": 49, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/d33fc7728b4aa3ee56aac4bf8f93edb4d16c507dea08e6fe4f772547e3c41dbd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/4d542c7c52d7dd3a6a97ec46deae73aee1a095193996ce78a7f3543fe66889e4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/ef405e24797035ac25e8fcc53215b94f301baf928c08e17b4896bf7371cc936a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6da01593-e133-55fe-8eb8-69be7217be58"}}
+{"definition": "neighborhood_attention_3d_h4_d64_kt11_ky11_kx11", "workload": {"axes": {"B": 4, "T": 121, "X": 128, "Y": 80}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/512b38651f67b074e3490efc032d9a878012da433e2c6a09ccf5aec4bb4b90cb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/6124f50af8107915feca4c0e9f38c1281be49e63c76b6bf341c817c374fe072d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h4_d64_kt11_ky11_kx11/2168ffede74cb7340bd792c6dd386c7ce130cff8e9e3464d23600e6867599b2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b188ec47-6dff-5a21-a6fb-1f0d4431ee7a"}}
diff --git a/workloads/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.jsonl b/workloads/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..88e2488a85a265fb1f6f517677c7975d32b379c5
--- /dev/null
+++ b/workloads/neighborhood_attention/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5.jsonl
@@ -0,0 +1,18 @@
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7b09c71ed021b4eee4b1acae2b0be13ec185f9c4a8bf493a1608b513405a7cb1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cf16c888513770ba3556910df8ac2fd03cca14dcc895f75c7d4270b0f2dd4923.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cb7f8de73340ef889f6f7ce6f0c56ff1291ef96880b54d09351310669f95de45.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "3eaae3b6-58d1-5ad4-94ba-ecf658aa5687"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/982fd582eeb5d044965f4202434b2c2a71fae218bdc2707d852e3cb89bd91e48.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/e9c7616f8320aaa6f720854c4bcbd92544578cd4d439f2dbca7392e29de2f9ba.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9f8c44559652875e6c9f9416a35195577709c915f2df3f2d3bba9498e7e3d3d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b573c4cd-8e41-5e92-af6a-362e7b16f0a1"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/42131e3a20e787b3f453edc1880609fdbb671deb33698c29a41b0dda8db27ae9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/346169bc0dcd5f579ed75b25f4284c3f5b49e01387419b8b9cde3cb2d7aa762c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c2fdca4a641213cf57c54b963e0f56e042db2719b4682ad0a1712c69a73355bf.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "95130464-5fd0-59c3-81b3-5759675a89f0"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/cc6c1bd811eb855e2dc3e3c2570b1e6190d6fc343910887e24b0611779d31137.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/43df4285fe22ceead84fe5b4ec111f98f8c92e3f41c942bc124671e0e34a4509.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1f893800031eb527f1bf8a2466a3fb8eb2b832f1c03cf3ba92293ea259cbe994.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d25c33ca-dcd8-58c5-b326-9a5e77c4d857"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4aceb1750a2f6179b8ce38d69a03bfc25ad735af4ea36c57ad480302e0987547.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7544e2bcfa8574207d6c0f292d21b2787b1ee72718d79736f22b5e61d3579466.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a29c67022f8b8027eaa1d0a773da5300cb7e5bb7e6650e165f84e71c8f1feee3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "65bd58a7-ae16-5b36-82e1-7c801c494fc3"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 1, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3188743a1c30b506e09708fda8c8086c9105b0b1a2d3c64700ff863f1842b160.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/b0b96451299326836450ebdce0a5ec228e44e3c2b896762ac03d008a69609531.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a778e4a9326c2a0464eb5426411d6a93e6cdc9064683db5fcfbec76a7fa0844c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "fdcc196f-3464-5db0-b64f-f2e29fe569fa"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1444cf5beaf689e2d5a6892c0e263711f4f2138db281ccdac2f033de5beae804.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/602cdbcb8df33cd2ac6d6f3744efe17e9c559561d1ad7a8febd834a9ebb1c2a5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6ee01f7306e95cd8cc0ee4624aa960829c402fcd74802b255ea98ea68e7a6ab2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7a5a1b75-2d8f-587c-842a-fdd539e9f60e"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/1ddeddaf64c922006361004eee4122caea1fe19e563ea8c68484e2a5c356e5a1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c40af9ce8e79af002d8d8ceb767650b8d88118a1c0c80132075017c5f9f94ebe.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/de35e1d879399c7f86f68f42c9e59f2e27a80352e4cb3d42bc6fb41c7fb5fc70.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "09b0007c-487c-5337-91cb-b6b9f153a218"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/a4c39f2fd24b76a65500dfb21b49017c8c407db4c0526eccab02203159ba8c51.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/fea1f01e1a4cc7497e6bfc4b4430c62843cc51340f8ddea42ae172f8009836d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/4451816a8fb4c69daa71d2488925a786faf12420ac58da2e195de0200ffa3b26.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1a60d4e5-bf8d-509a-aa67-96eb03bd8e7e"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3b97efc0c7e6e163ca4f81dcf9d1fe65facf5b663fab1d6ff96422c39e9a8137.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9b7ccca7fefa1e4b6da881b1ef72f35923e5164dc826b6a9a21e53cc90d85340.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8f337574d6086ba92a114a619a94693d57b68d2b8f2035027b07ce927ad80aad.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6ed37cda-b6d1-57c6-bbd2-5ef073b679e1"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d570ae295a9a879f1be6f36dfac7e658b42ceb814501ed8b778ec870a69fd97f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/090fa7e62e142c4f55830cea4d9b13b7ae7f8c94cde112f6ea4ff16135327ec5.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/ce140fe06c5d6dc83a8199541495ae94ae640223f20d3a3f8476886758e91c3a.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c33b8ec1-0039-5090-816e-43d244a31e5a"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 2, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/22bb3fcc2e673f4526439b08176b2270317b7089ff0d7c1201e9beffbed6061c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/60154d4d3589527f266e94603e8c40ea6b510cd000f10af21573ab8a61492ecf.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/be8749ef5915f2bcebdb2e2281755cdd77f0be2f1cbc5e392c2e27841e1c8579.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6465f201-d904-5e1c-abcb-a2b6b4061ecf"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 17, "X": 48, "Y": 28}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/2514447930a6a3db65bbd7098dfaf7082e44920225cde2f5d1833b27db917f06.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6a190d9de30310878542b6b32c37a9787fbb7ab1e509e568ea7c7f63dd672671.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/8b2231abd2d7f38f4e0fe3178a53d48f34c0b23720c9930e20e47873c92bd7f6.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "14c7aef4-8e42-5a1b-8f42-16146a940901"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 33, "X": 96, "Y": 56}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/d587ed9d02c3654f24869e933084a7e446d444e318574b4ce5bc0cf2047f1beb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/73e4c436b9a73d31b14626f7a1fd089a57d081c90dc604b0636e85ab3105ee6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/91690ed88888fcd0c697c1165a2d984daa7f83c5d2c22d9104cb190ece50ba89.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0253ef4d-943e-5839-9cb0-8d384be40c67"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 17, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/0445ae182e7855112e428c41441fb62d383ca80c3e9e8f12c1c20ca00718a04b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6fda52652daebe9b988c9f401c852dc50b0c2c5b36a7c957e2de95fbeee3f059.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/14af1d4927000d1902e2f4347f5e8deb92fd5cf0b98ade6015e7237f06f15681.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ece8abe8-c6b6-55f8-8585-e455a076854b"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 33, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/9fb12b166429a3924cf88e1b3c59d4f2c07efb2d1e4c8ccdd7b66a639b5bb46b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c6eb51881525b4266f6cdccaf31ce2e4098d0e4e797c53c50a1035c01185965f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/183d15bd27a5b9bc2c2b496a4ca853cbac6509850cb615f311b3427ba1dfdfc6.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "abe8cf54-16a2-5f37-b119-71880cdf77d5"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 35, "X": 32, "Y": 20}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/528dff635c8de2da293cbf52f95420f91c112602f39518dea9365ac63a96156d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/6277dfbc929bc956086c093395a0c8f59c7db3d12173ba99044d11381db842ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/7329161b346f9bce7ecdfea5517f10a6fa3dd389eeae0dabdc1d572974ca27a3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "9866c54d-a157-5fdd-89d1-84af0fb714e6"}}
+{"definition": "neighborhood_attention_3d_h8_d64_kt3_ky5_kx5", "workload": {"axes": {"B": 4, "T": 69, "X": 64, "Y": 40}, "inputs": {"K": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/57a9f23d1582396ca82b8d0205ab723f795d263d99abf2ba2ddd86f5f601fbde.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/3f6ab3c388f9e692304a3aea4a9a1c5385f5fd12224c19764546ea9501648b2b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "V": {"path": "blob/workloads/neighborhood_attention_3d_h8_d64_kt3_ky5_kx5/c62cca8d265d87edb82e3855fd01939036a843f27a10de0f67dd8f41ce13cce5.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "e575ca18-1da8-5e40-8439-66d5e614c4c5"}}
diff --git a/workloads/sparse_mla/sparse_mla_prefill_h64_d512.jsonl b/workloads/sparse_mla/sparse_mla_prefill_h64_d512.jsonl
deleted file mode 100644
index dc8569def26a0dcb5be6b38995274d1e5e6c13fb..0000000000000000000000000000000000000000
--- a/workloads/sparse_mla/sparse_mla_prefill_h64_d512.jsonl
+++ /dev/null
@@ -1,30 +0,0 @@
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 128, "Nkv": 128, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2944, "Nkv": 2944, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 448, "Nkv": 575, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1c89357-aa68-53ac-b13f-62d900007022"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 3456, "Nkv": 3583, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2432, "Nkv": 2559, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a834433-bbcc-5122-b239-be8569f77d02"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 4096, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 4223, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4032, "Nkv": 4286, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 1984, "Nkv": 2238, "TopK": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 128, "Nkv": 192, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 128, "Nkv": 256, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2944, "Nkv": 4416, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2944, "Nkv": 5888, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "52560299-af35-5351-a4f0-cdcf97141e87"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 448, "Nkv": 71583, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 448, "Nkv": 142591, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 277247, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 550271, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 3456, "Nkv": 278335, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 3456, "Nkv": 553087, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2432, "Nkv": 497471, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 2432, "Nkv": 992383, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 6144, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 8192, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7fd83f55-ee1c-5bec-b886-047096621576"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 8319, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4096, "Nkv": 12415, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4032, "Nkv": 39070, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 4032, "Nkv": 73854, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 1984, "Nkv": 37662, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797"}}
-{"definition": "sparse_mla_prefill_h64_d512", "workload": {"axes": {"M": 1984, "Nkv": 73086, "TopK": 640}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "43e73df3-0172-583e-944e-789dff99b8b6"}}
diff --git a/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k128.jsonl b/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k128.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6e416b750bf784aa583dcc9b5eb04b474f7a6f27
--- /dev/null
+++ b/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k128.jsonl
@@ -0,0 +1,10 @@
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 128, "Nkv": 128}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/572599b383b42fb14c09ecb54c65aa55a78da039df2a1e17b899e67470c3b93b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/e260cdbaaf395c0eef8f34946ed46926cae26f357249efa7dc85b9861bc710a8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/d044c27a0ebeb9e63a4c8e8ad610e2135b7d4acb868179b98a84a02ec05f6993.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/64a2c7d30f0846c08ed9643a2d4f6c92b178dac0951e97627cce17aea69aa268.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "63072e3a-2bf9-5eed-bc07-c48f323b15d4"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 2944, "Nkv": 2944}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/be1b48af452d49f50f15576a8afe71475513b6ea7efbb9bab6d289e181ace151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/4de7e30b99d0b89f616d6da14bde1aab5c16b3268e74ce4a3e72bb80c49a3503.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/593be81ca82e4187fa1a94b205ffa86d79c27019575d1dda033e1cca7c91f8ed.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/c256c9790e86022b3954dd7baf3396a0442e91ca06474e51a9d90f15fbe39bb3.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "ec18845f-e1bb-5eba-9649-3d90d470ce9b"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 448, "Nkv": 575}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/b154b9425f2e4b6ca12055d4a9bfe62dd9b1ab69158310f559ec88c802aa2977.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/e9ca9b862dff67ebd39109ad3bae467e1aa908dcb92a0cf7da4f4d9c74331d07.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f20167148c7e3559ea97e22d019300cca1e0306a9984980770f542b4ac889f6a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/6caca87bc0190236a5e86277fc2d76883f09f7cbcaf9c8b233c128db373c7041.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1602ca9-fae0-56ef-9bb2-a408fe19a22b"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 4096, "Nkv": 4223}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/8b451ff76b639ecbdcc6e0a4b277a31318762fed24d149604f7117723edbcaa2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/31c8dfe2a0e98ce39a76e48d1c050fb861cae7a4008bc2a02b9070e03b5fcf2d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d1c89357-aa68-53ac-b13f-62d900007022"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 3456, "Nkv": 3583}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/e0e8bbbec9416f6f7d68f50d1c3aa9b1513ba4766964b239f945606f2bd54222.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/b4a4e7df9e9af433d6016f32596942993a74564e916c0d946f99461a7966b1c9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/bace4df66694e822f673471b6588e15500531cd9f5b1e0880a3983024627a577.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/0ce973c67b80e92bc52c94e699a4b8745c41d6039162d3b93bc4609625626343.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c315bd09-f7c1-5174-bed2-66b2e2b8d7be"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 2432, "Nkv": 2559}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/2719ad4a6672b4960eddab778a80f0dc8b86e2ff707e9d094e950c3e9dbaac3e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/5534f37f6c88693185ba982c7801a65af22f7802a0476115fd60f72793bc7599.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/b8dc8f943d7961731390ecc664ba1e455cecb88468f3821914b9cf91f29bfe7e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/48bc5365bf0e852045c2a206d26412832c8914caec9643b329e9d44994e5464d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "8a834433-bbcc-5122-b239-be8569f77d02"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 4096, "Nkv": 4096}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f11c0db5114b1f5f0687820db49b3d76c86691d89108e4b037207441f414fffc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/e3c224642165f2756eb5d6023e85ac808aa3082544dc2ce8cc2558554d51b161.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/0712464b6cbde0cb38db50717f740f6b02f87ba28b8da7bff513c60a3c97613f.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/1ea1907b559fae5693724aecc9dedfaad94355ac1c8bb027f6a49394b13cbbda.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5bb71426-9b5d-5301-a44d-b6b4afc5a924"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 4096, "Nkv": 4223}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/74c96c7c7a6bd1496db00413426191c1d64f38e25ea151fc5ad0dfb94182a462.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/7196bdaac49625a14b4772405f5db100e6d6ff5b6e6002cb19f342b6abe19b17.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/a898d898c8c18bfd0b5548135f8d7034b25ec0a5ab8bd1d700bfc59762041151.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f467609e12c1a8c05c6a584e11841423c0c6eac93608075909023ac37e7bbcc7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f475abca-84d5-58f9-a8b0-ee05530a4fcc"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 4032, "Nkv": 4286}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/4758a25a55c2809920f4a72898c9b8199288373f2b3017e9b996fb5a33c89188.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/7cb74378b7f9e9620ebbbad3357e7cb0a2bf381dbc2da254235cf1940f9e3b35.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/824f9dd1407658950fe06ad814ed215d79cfb1e4006847a0be1b8b7aacf272bd.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/2a6a394516fb740837157690cd708280b781c4f48ed7c11026a5b6d9859cfc89.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c6c6e0c9-61e2-505e-831a-fef983c9dad9"}}
+{"definition": "sparse_mla_prefill_h64_d512_k128", "workload": {"axes": {"M": 1984, "Nkv": 2238}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/a32e8fa49824fdfd91505c0d1de90922950bb7ab94be8966ba50005ab78a2063.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/87a587dcdf20532eb539481b9f7e2769e998f2e60985c2a9c0b0f359b10bef27.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/f493e5837d61b98771979407ffa0687184c5a22ba90f19d6bea57dda5ff4bc4a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/498113893435eaca7331fa8a907f2293f45871777746e6cfeb832b6fc13f4e7b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k128/976b20fd85a83a08dd8f0603504312c5be07e479271a4470c00d7114d3bd219c.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "aa1f4b6e-c84d-5723-8007-090ffa76004f"}}
diff --git a/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k640.jsonl b/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k640.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..042a5eee8da867bc5d8a96aa38cc0805c5d14756
--- /dev/null
+++ b/workloads/sparse_mla/sparse_mla_prefill_h64_d512_k640.jsonl
@@ -0,0 +1,20 @@
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 128, "Nkv": 192}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/a901e15535fd0664de38cb7aa51e5df02605c1dcd2b9dd23a7cbe107ac64a50e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/fecdcfe9dae23eed1d55fc584f754a69de387af1c4b130c7e683b69eed4d1cad.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/abd5108b25b88231d338fc44f9a8f641c3d41943d068cc0e2c5f146e8c3f5570.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/2d449d76961b0da6b1be34633d8ef0c0ee186eacf342ceebebfc153e2cef294e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d9165058-a73b-5e77-8e61-5f92a8458b70"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 128, "Nkv": 256}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b4a8b5bc9719f755b63925ddfaa3decbeeef6e1089d41b52ccfa950971b0990d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/737eec3da7d5320f8e6b7daea4833c92b273ebafd991733d09714a4d7636d1e3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/979a14b0cf096acc1a400d27a96d01fb0f8704d833cb77b64cdfbcb037364e66.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b987b1867c4873d19dcf92622dee76a9c7b4f534345cb3adf5becf2851727107.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "546bcf0e-32bd-5dd8-9ea9-14c58e6a52c9"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 2944, "Nkv": 4416}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/057ee2578503f838e5a1c609b588f2f55247fd7c524dc7099e79f250b699b83b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/58f9c4267f4487e3576879429d1fd138858765d6825b7cdaa7c3d815f0a1fa52.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8c711fecae3bb34b01f969d7d332b9c9c1d7b9eb6a28bcd11f1d01d0809ce3b1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/25fa8d12e7a7ce1bc0df07db4f1cbfb8f2c7e5db3e69128a541b209df450d06e.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6a0463ea-da74-5e52-a6f4-18548159142a"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 2944, "Nkv": 5888}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/94fbb6f81eef550921e31981fe0cce84a32856a457024f553bc78aac7d851fd1.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/576a1064df10db25df0052bd835e109d474b095221ecf38c498bf582d347e2d2.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/00fbbfb0c5b758642b6b40368d1038dcfc2c0ff0715f3ffb6b9f89313d416608.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8353e51f54c5758c67005def018c9f2792cf1ab6a7efd12006f7b87c2dbaf3b4.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "52560299-af35-5351-a4f0-cdcf97141e87"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 448, "Nkv": 71583}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/cf939862e1812342922374c1481c081456705ba78e0a6f47f7ec04c533e86afb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/feadc71d38dc74c1ca2d3512f163a6de0d9aa4c4b21b8474eb5508400f21b995.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/e1790d69193dd09904e1e09385d6fb7328d89d8db1fe05b0fc6e36018d2016d3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7d138df7-8e65-541a-9ad1-d572d5131b5b"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 448, "Nkv": 142591}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8cee08b1837e9a881189431d23a86c585bc8093f25b5834a97276c5169bc7beb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/4cd3eb3df57b6eb09c3337038ec8f07d2cf7815b55c269eb033006139fd4f5c4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/955aa5aeaa3ba2d3d795a01d4f843229228122efba71a605a218e174b917415e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/ddaaeb53984f937526b6d7503a33e439ebadb23d4bc4f857692c4c1045001141.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "495000a6-c8b0-5057-99f4-88ed4a96a4a1"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 277247}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/58221ef18bd44a5026ac8d71cecc63b31e81d9fb8aec0308d8866098ea695e2a.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/46d18e627098be763b7ab264b3e153ad26631ad42d04da41dfe86391c1d5e827.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/17282df5f7365ee75ce42f732d2ef17119bfa87e4426aab4afb67e763f16100e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "1452f1b3-6d7e-5490-ba5d-ca2276922417"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 550271}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/26481c5784d502d23af795cdfd06d40c9542502b19f2847c093aeff3b8a1da78.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d464004dfb29fefdfe6ba4d2ec61bb192a294a73c1eb7f843bed1206892a431d.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/3672fc0d257c41f6c1bf8251ce43ed44ac095a0c719e7cf7c6c622f6ac8b8c36.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "c0332d0e-dddb-5656-a8c7-0ced515dbbfa"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 3456, "Nkv": 278335}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/341752259a22e2f21b3512219cb4744baf746cf65a6467e10f4e56dd327c85a0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/f0e39d82c7a6ae0aa5ddb96520faf5f2c272e1df03a6e11bb24ff33365e22d24.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/a748b3fedd459fa6df11fe4dd8628d14f4b2fa2843127e19dad46ca382151c4c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "d22bc34f-f49e-5883-a140-f1c5f6bbb8ed"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 3456, "Nkv": 553087}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/5b176dc0bab264a10266e5dda50e4efd4657c7820518cc23f9181e3c2fa69a1e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/97ed1ac19a23b9c7cebb98b2f85e8baf3bd003b94b8f07d047b4f210d5f4513e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/01544883ad6db9de47733c90539d4f1dbb68b9f626448174026b2c144bf9aec3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/4963acd377179c61e9a992a263060f889fea16fb90fe6c7b3393a69b96acc25d.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "5e5ad68f-a0a3-56e6-a0c3-065765feb8b2"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 2432, "Nkv": 497471}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/51e99e6d4b7097af19f3c09a3b1e34ce4b5f73583b813c4bd8250882a85ac4cc.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/2878582148bbc3d3d9ac28b7591c750d9610bbdc5acd0ad48e830908e444457e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/eb067f526089920ef7dc6f68534e0fb4f7f75c97b579ca4dc1ae3293aea0b440.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "762c5a7d-7dd9-5e91-a328-06c62d9f471a"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 2432, "Nkv": 992383}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/75cd3e97179d49fa1caf4d0a797294e058012e7f3f8d10c1d49bbd99139f4b22.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29e119b7d879320c33cc96182fb93ad56ae4953f047d0fc81ccaccc572d9bf54.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d5e24eab060743d74d666c401695e085ad3ba969cd9efff61148f8c0675f6afb.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/23f39f23ffcea14fd207e8859b62412a8788e61cf4dfec882f96b73b5fb24ed8.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "42f3b7ee-5af7-5dd1-bf58-71493057d3f5"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 6144}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/c61d52780bc8da46119aee1d8b84851501546e789f018e01768cbaf690cd3b83.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/c1395a053462f6ab76c72a32df3e0bab06806b731bbedfc5f0fd7abace8c1194.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/f15088bf48b179b2ddbfd54182c8f32738077002f2f24a38d75f5e8cad06d292.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/a6a4784ac41c81e4f5ea8d65568857bf2a2db05f1da82e46b850172aa3090913.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b57f47c5-314a-57db-9233-9b1470763ce5"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 8192}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/2be37865a899e6ca9304320a9f730262eeb546b49b26a193d117c99cc319fdb0.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/37f3dbd8c15c7f76babd09a8e0560e247d4df5beeb4ef0b2cd8cacd6e58ee655.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/bfe313125005154deaf12f3017d90430d35672a2697b6684440880dfedc93825.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/523eba4889d1aafd922cd9bc8e578eb15ac7f50ec43fbe1980c9d4649085c8d7.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "7fd83f55-ee1c-5bec-b886-047096621576"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 8319}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/af3924a1b23010bb7baa9f4e440bc835ba72f1634d5eb9936c74c3d7bee4c091.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d9fdc1ad51b41fb91d7d96ccf9893b43a07ef32787d4d12e1245feb63a43c5ef.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/750c1b955a6de1acfdf79a33e1893e63492e903e6d656c38ef650f443cfe4595.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "f0c8462c-5541-51c8-90c2-52686dd93402"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4096, "Nkv": 12415}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8ae828c8362c2def26569d4355f8851a9d73ec1e1fb7d0d8970ffbc33e160949.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/925d7a5be741a53bb5260c1d370f1422f8e2e4d9bd768e98c93c1a20994c07d4.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/8960cd9c7b1d54f28978bcc41f2fa15dba382a6c5e0d446f32409cea550d7347.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/1035919ebf6bc23761079c79c0193bb6cb838d40a12f0ecec2d9f835b60b33b2.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "b1ee31d7-328c-59af-a60c-189c90e2d248"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4032, "Nkv": 39070}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/21eb9fd5458d0fc98550d85beed7a16d6b9ce198017f8d291175d5584f3f5d6e.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/9bac59d9dbbad96ad5ca74ad0979e26c9cc8a3a1637b5dd1f36eab68693ea27c.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/21fb0b30717b51baa8ae0a259491d3b9cb463f0faf9fd192a8576ffde62fe181.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "2c6e4648-5da2-580e-a26e-46e6f300b0a5"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 4032, "Nkv": 73854}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/436235f97e9de45c4bb50b5beca48e9c46b9f7b911ad9c3d1f59f0d1eebef738.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/402b0097867c9977f248deaa8256e7e747e79730e3b4332777acf564ded32882.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/729302fbb0995b5eb282b1a740a49f9153c8f5f048ace8a6b696568a0d2385b3.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/271f7a8c7215fa24800e058b99437d173bfed890c5626dc8b1c25ff0796a52c1.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "6385d1ff-f172-5c14-b73d-1c2b5943f045"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 1984, "Nkv": 37662}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/0b7cab7cea14234da8c9b4b9247623a309bd83932a46ce5fb65d1d9dc1743e67.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/106f7147bf098f471252b712f8c70a0e7c11f9d90b0c783144b2716a3adb55be.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/b779ffba7d911e6831944b33a1b6cc51a98a13ff86bdcd5e66a4fec25166b28b.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/0356da780595567c481410df5cf0ad8eb300a3243049ba8b11402072b5ff4982.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "0b988a7a-2b52-5f49-8c99-eaa373baf797"}}
+{"definition": "sparse_mla_prefill_h64_d512_k640", "workload": {"axes": {"M": 1984, "Nkv": 73086}, "inputs": {"KV": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/82110ca467f9fdc3535b7c3f0287f72da4ebc433b747bc507a4ce7c9d6eb7504.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "Q": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/714a89a405cf4b2fbebe14da1ba238abfc6cadf4d7312f91eef63ae9b110a8b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "attn_sink": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/29acf81ebc976007dd888f95d70500a32c1e7c767c7b889ab230f5b0528b52b8.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "indices": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/d494837d9eccd3ff71a0f0ab528cfd87e30ec0fe8d0c49e786004cf8f1452bc9.safetensors", "tensor_key": "tensor", "type": "safetensors"}, "sm_scale": {"type": "scalar", "value": 0.04419417382415922}, "topk_length": {"path": "blob/workloads/sparse_mla_prefill_h64_d512_k640/c832ab4b31156f1af50d687d941df1512c6ae3ef271eac80c62809e0407bb410.safetensors", "tensor_key": "tensor", "type": "safetensors"}}, "uuid": "43e73df3-0172-583e-944e-789dff99b8b6"}}